updated

Signed-off-by: Robert Shaw <robshaw@redhat.com>

.buildkite/check-wheel-size.py

@@ -5,11 +5,11 @@ import os
 import sys
 import zipfile
 
-# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 500 MiB
+# Read the VLLM_MAX_SIZE_MB environment variable, defaulting to 450 MiB
 # Note that we have 800 MiB quota, please use it wisely.
 # See https://github.com/pypi/support/issues/6326 .
 # Please also sync the value with the one in Dockerfile.
-VLLM_MAX_SIZE_MB = int(os.environ.get("VLLM_MAX_SIZE_MB", 500))
+VLLM_MAX_SIZE_MB = int(os.environ.get("VLLM_MAX_SIZE_MB", 450))
 
 
 def print_top_10_largest_files(zip_file):

.buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-QQQ.yaml

@@ -1,12 +0,0 @@
-# For vllm script, with -t option (tensor parallel size).
-# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m HandH1998/QQQ-Llama-3-8b-g128 -b 32 -l 1000 -f 5 -t 1
-model_name: "HandH1998/QQQ-Llama-3-8b-g128"
-tasks:
-- name: "gsm8k"
-  metrics:
-  - name: "exact_match,strict-match"
-    value: 0.419
-  - name: "exact_match,flexible-extract"
-    value: 0.416
-limit: 1000
-num_fewshot: 5

.buildkite/lm-eval-harness/configs/Meta-Llama-4-Maverick-17B-128E-Instruct-FP8-MM.yaml

@@ -1,12 +0,0 @@
-# For hf script, without -t option (tensor parallel size).
-# bash .buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh -m meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 -l 100 -t 8
-model_name: "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
-backend: "vllm-vlm"
-tasks:
-- name: "chartqa"
-  metrics:
-  - name: "relaxed_accuracy,none"
-    # TODO(zhewenl): model card is 0.90, but the actual score is 0.80.
-    value: 0.80
-limit: 100
-num_fewshot: 0

.buildkite/lm-eval-harness/configs/Meta-Llama-4-Maverick-17B-128E-Instruct-FP8.yaml

@@ -1,10 +0,0 @@
-# For hf script, without -t option (tensor parallel size).
-# bash .buildkite/lm-eval-harness/run-lm-eval-mmlupro-vllm-baseline.sh -m meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 -l 250 -t 8 -f 5
-model_name: "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
-tasks:
-- name: "mmlu_pro"
-  metrics:
-  - name: "exact_match,custom-extract"
-    value: 0.80
-limit: 250 # will run on 250 * 14 subjects = 3500 samples
-num_fewshot: 5

.buildkite/lm-eval-harness/configs/Qwen2.5-VL-3B-Instruct-FP8-dynamic.yaml

@@ -1,5 +1,4 @@
-# For vllm script, with -t option (tensor parallel size)
-# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic -l 1319 -t 1
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic -b auto -l 1319 -f 5 -t 1
 model_name: "RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic"
 tasks:
 - name: "gsm8k"

.buildkite/lm-eval-harness/configs/Qwen2.5-VL-7B-Instruct.yaml

@@ -1,12 +0,0 @@
-# For vllm script, with -t option (tensor parallel size).
-# bash .buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh -m Qwen/Qwen2.5-VL-7B-Instruct -l 2500 -t 1
-
-model_name: "Qwen/Qwen2.5-VL-7B-Instruct"
-backend: "vllm-vlm"
-tasks:
-- name: "chartqa"
-  metrics:
-  - name: "relaxed_accuracy,none"
-    value: 0.855
-limit: 2500
-num_fewshot: 0

.buildkite/lm-eval-harness/configs/Qwen3-235B-A22B-Instruct-2507-FP8.yaml

@@ -1,14 +0,0 @@
-model_name: "Qwen/Qwen3-235B-A22B-Instruct-2507-FP8"
-tasks:
-  - name: "mmlu_pro"
-    metrics:
-      - name: "exact_match,custom-extract"
-        value: 0.82
-limit: 250 # will run on 250 * 14 subjects = 3500 samples
-num_fewshot: 5
-enforce_eager: false # we use false to speed up the eval process
-kv_cache_dtype: fp8 # we use fp8 to speed up the eval process
-max_model_len: 40960
-apply_chat_template: true
-fewshot_as_multiturn: true
-gen_kwargs: "temperature=0,top_p=1,top_k=0,max_gen_toks=5632,until=<|ENDANSWER|>"

.buildkite/lm-eval-harness/configs/models-large-hopper.txt

@@ -1 +0,0 @@
-Qwen3-235B-A22B-Instruct-2507-FP8.yaml

.buildkite/lm-eval-harness/configs/models-mm-large-h100.txt

@@ -1 +0,0 @@
-Meta-Llama-4-Maverick-17B-128E-Instruct-FP8-MM.yaml

.buildkite/lm-eval-harness/configs/models-mm-small.txt

@@ -1 +0,0 @@
-Qwen2.5-VL-7B-Instruct.yaml

.buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh

@@ -1,44 +0,0 @@
-#!/bin/bash
-# We can use this script to compute baseline accuracy on chartqa for vllm.
-#
-# Make sure you have lm-eval-harness installed:
-#   pip install lm-eval==0.4.9
-
-usage() {
-    echo``
-    echo "Runs lm eval harness on ChartQA using multimodal vllm."
-    echo "This pathway is intended to be used to create baselines for "
-    echo "our correctness tests in vllm's CI."
-    echo
-    echo "usage: ${0} <options>"
-    echo
-    echo "  -m    - huggingface stub or local directory of the model"
-    echo "  -l    - limit number of samples to run"
-    echo "  -t    - tensor parallel size to run at"
-    echo
-}
-
-while getopts "m:l:t:" OPT; do
-  case ${OPT} in
-    m ) 
-        MODEL="$OPTARG"
-        ;;
-    l ) 
-        LIMIT="$OPTARG"
-        ;;
-    t ) 
-        TP_SIZE="$OPTARG"
-        ;;
-    \? ) 
-        usage
-        exit 1
-        ;;
-  esac
-done
-
-lm_eval --model vllm-vlm \
-  --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE" \
-  --tasks chartqa \
-  --batch_size auto \
-  --apply_chat_template \
-  --limit $LIMIT

.buildkite/lm-eval-harness/run-lm-eval-mmlupro-vllm-baseline.sh

@@ -1,50 +0,0 @@
-#!/bin/bash
-# We can use this script to compute baseline accuracy on MMLUPRO for vllm.
-# We use this for fp8, which HF does not support.
-#
-# Make sure you have lm-eval-harness installed:
-#   pip install git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d#egg=lm-eval[api]
-
-usage() {
-    echo``
-    echo "Runs lm eval harness on MMLU Pro using huggingface transformers."
-    echo "This pathway is intended to be used to create baselines for "
-    echo "our automated nm-test-accuracy workflow"
-    echo
-    echo "usage: ${0} <options>"
-    echo
-    echo "  -m    - huggingface stub or local directory of the model"
-    echo "  -l    - limit number of samples to run"
-    echo "  -f    - number of fewshot samples to use"
-    echo "  -t    - tensor parallel size to run at"
-    echo
-}
-
-while getopts "m:b:l:f:t:" OPT; do
-  case ${OPT} in
-    m )
-        MODEL="$OPTARG"
-        ;;
-    b )
-        BATCH_SIZE="$OPTARG"
-        ;;
-    l )
-        LIMIT="$OPTARG"
-        ;;
-    f )
-        FEWSHOT="$OPTARG"
-        ;;
-    t )
-        TP_SIZE="$OPTARG"
-        ;;
-    \? )
-        usage
-        exit 1
-        ;;
-  esac
-done
-
-lm_eval --model vllm \
-  --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,add_bos_token=true,trust_remote_code=true,max_model_len=4096" \
-  --tasks mmlu_pro --num_fewshot "$FEWSHOT" --limit "$LIMIT" \
-  --batch_size auto

.buildkite/lm-eval-harness/test_lm_eval_correctness.py

@@ -19,35 +19,21 @@ RTOL = 0.08
 def launch_lm_eval(eval_config, tp_size):
     trust_remote_code = eval_config.get("trust_remote_code", False)
     max_model_len = eval_config.get("max_model_len", 4096)
-    batch_size = eval_config.get("batch_size", "auto")
-    backend = eval_config.get("backend", "vllm")
-    enforce_eager = eval_config.get("enforce_eager", "true")
-    kv_cache_dtype = eval_config.get("kv_cache_dtype", "auto")
     model_args = (
         f"pretrained={eval_config['model_name']},"
         f"tensor_parallel_size={tp_size},"
-        f"enforce_eager={enforce_eager},"
-        f"kv_cache_dtype={kv_cache_dtype},"
+        f"enforce_eager=true,"
         f"add_bos_token=true,"
         f"trust_remote_code={trust_remote_code},"
-        f"max_model_len={max_model_len},"
+        f"max_model_len={max_model_len}"
     )
     results = lm_eval.simple_evaluate(
-        model=backend,
+        model="vllm",
         model_args=model_args,
         tasks=[task["name"] for task in eval_config["tasks"]],
         num_fewshot=eval_config["num_fewshot"],
         limit=eval_config["limit"],
-        # TODO(yeq): using chat template w/ fewshot_as_multiturn is supposed help
-        # text models. however, this is regressing measured strict-match for
-        # existing text models in CI, so only apply it for mm, or explicitly set
-        apply_chat_template=eval_config.get(
-            "apply_chat_template", backend == "vllm-vlm"
-        ),
-        fewshot_as_multiturn=eval_config.get("fewshot_as_multiturn", False),
-        # Forward decoding and early-stop controls (e.g., max_gen_toks, until=...)
-        gen_kwargs=eval_config.get("gen_kwargs"),
-        batch_size=batch_size,
+        batch_size="auto",
     )
     return results
 

.buildkite/nightly-benchmarks/README.md

@@ -2,23 +2,40 @@
 
 ## Introduction
 
-This directory contains a benchmarking suite for **developers** to run locally and gain clarity on whether their PR improves/degrades vllm's performance.
-vLLM also maintains a continuous performance benchmark under [perf.vllm.ai](https://perf.vllm.ai/), hosted under PyTorch CI HUD.
+This directory contains two sets of benchmark for vllm.
+
+- Performance benchmark: benchmark vllm's performance under various workload, for **developers** to gain clarity on whether their PR improves/degrades vllm's performance
+- Nightly benchmark: compare vllm's performance against alternatives (tgi, trt-llm and lmdeploy), for **the public** to know when to choose vllm.
+
+See [vLLM performance dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm) for the latest performance benchmark results and [vLLM GitHub README](https://github.com/vllm-project/vllm/blob/main/README.md) for latest nightly benchmark results.
 
 ## Performance benchmark quick overview
 
-**Benchmarking Coverage**: latency, throughput and fix-qps serving on B200, A100, H100 and Intel® Xeon® Processors, with different models.
+**Benchmarking Coverage**: latency, throughput and fix-qps serving on A100 (the support for FP8 benchmark on H100 is coming!) and Intel® Xeon® Processors, with different models.
 
 **Benchmarking Duration**: about 1hr.
 
 **For benchmarking developers**: please try your best to constraint the duration of benchmarking to about 1 hr so that it won't take forever to run.
 
+## Nightly benchmark quick overview
+
+**Benchmarking Coverage**: Fix-qps serving on A100 (the support for FP8 benchmark on H100 is coming!) on Llama-3 8B, 70B and Mixtral 8x7B.
+
+**Benchmarking engines**: vllm, TGI, trt-llm and lmdeploy.
+
+**Benchmarking Duration**: about 3.5hrs.
+
 ## Trigger the benchmark
 
-The benchmark needs to be triggered manually:
+Performance benchmark will be triggered when:
+
+- A PR being merged into vllm.
+- Every commit for those PRs with `perf-benchmarks` label AND `ready` label.
+
+Manually Trigger the benchmark
 
 ```bash
-bash .buildkite/performance-benchmarks/scripts/run-performance-benchmarks.sh
+bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
 ```
 
 Runtime environment variables:
@@ -30,6 +47,10 @@ Runtime environment variables:
 - `REMOTE_HOST`: IP for the remote vLLM service to benchmark. Default value is empty string.
 - `REMOTE_PORT`: Port for the remote vLLM service to benchmark. Default value is empty string.
 
+Nightly benchmark will be triggered when:
+
+- Every commit for those PRs with `perf-benchmarks` label and `nightly-benchmarks` label.
+
 ## Performance benchmark details
 
 See [performance-benchmarks-descriptions.md](performance-benchmarks-descriptions.md) for detailed descriptions, and use `tests/latency-tests.json`, `tests/throughput-tests.json`, `tests/serving-tests.json` to configure the test cases.
@@ -131,3 +152,26 @@ Here is an example using the script to compare result_a and result_b with Model,
 A comparison diagram will be generated below the table.
 Here is an example to compare between 96c/results_gnr_96c_091_tp2pp3 and 128c/results_gnr_128c_091_tp2pp3
 <img width="1886" height="828" alt="image" src="https://github.com/user-attachments/assets/c02a43ef-25d0-4fd6-90e5-2169a28682dd" />
+
+## Nightly test details
+
+See [nightly-descriptions.md](nightly-descriptions.md) for the detailed description on test workload, models and docker containers of benchmarking other llm engines.
+
+### Workflow
+
+- The [nightly-pipeline.yaml](nightly-pipeline.yaml) specifies the docker containers for different LLM serving engines.
+- Inside each container, we run [scripts/run-nightly-benchmarks.sh](scripts/run-nightly-benchmarks.sh), which will probe the serving engine of the current container.
+- The `scripts/run-nightly-benchmarks.sh` will parse the workload described in [nightly-tests.json](tests/nightly-tests.json) and launch the right benchmark for the specified serving engine via `scripts/launch-server.sh`.
+- At last, we run [scripts/summary-nightly-results.py](scripts/summary-nightly-results.py) to collect and plot the final benchmarking results, and update the results to buildkite.
+
+### Nightly tests
+
+In [nightly-tests.json](tests/nightly-tests.json), we include the command line arguments for benchmarking commands, together with the benchmarking test cases. The format is highly similar to performance benchmark.
+
+### Docker containers
+
+The docker containers for benchmarking are specified in `nightly-pipeline.yaml`.
+
+WARNING: the docker versions are HARD-CODED and SHOULD BE ALIGNED WITH `nightly-descriptions.md`. The docker versions need to be hard-coded as there are several version-specific bug fixes inside `scripts/run-nightly-benchmarks.sh` and `scripts/launch-server.sh`.
+
+WARNING: populating `trt-llm` to latest version is not easy, as it requires updating several protobuf files in [tensorrt-demo](https://github.com/neuralmagic/tensorrt-demo.git).

.buildkite/nightly-benchmarks/benchmark-pipeline.yaml

@@ -0,0 +1,184 @@
+steps:
+  - label: "Wait for container to be ready"
+    key: wait-for-container-image
+    agents:
+      queue: A100
+    plugins:
+    - kubernetes:
+        podSpec:
+          containers:
+          - image: badouralix/curl-jq
+            command:
+            - sh .buildkite/nightly-benchmarks/scripts/wait-for-image.sh
+  - label: "Cleanup H100"
+    agents:
+      queue: H100
+    depends_on: ~
+    command: docker system prune -a --volumes --force
+  
+  - label: "A100"
+    # skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
+    agents:
+      queue: A100
+    depends_on: wait-for-container-image
+    if: build.branch == "main"
+    plugins:
+    - kubernetes:
+        podSpec:
+          priorityClassName: perf-benchmark
+          containers:
+          - image: public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT
+            command:
+            - bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+            resources:
+              limits:
+                nvidia.com/gpu: 8
+            volumeMounts:
+            - name: devshm
+              mountPath: /dev/shm
+            env:
+            - name: VLLM_USAGE_SOURCE
+              value: ci-test
+            - name: HF_TOKEN
+              valueFrom:
+                secretKeyRef:
+                  name: hf-token-secret
+                  key: token
+          nodeSelector:
+            nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
+          volumes:
+          - name: devshm
+            emptyDir:
+              medium: Memory
+
+  - label: "H200"
+    # skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
+    agents:
+      queue: H200
+    depends_on: wait-for-container-image
+    if: build.branch == "main"
+    plugins:
+    - docker#v5.12.0:
+        image: public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT
+        command:
+        - bash
+        - .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+        mount-buildkite-agent: true
+        propagate-environment: true
+        ipc: host
+        gpus: 4,5,6,7
+        volumes:
+          - /data/benchmark-hf-cache:/root/.cache/huggingface
+        environment:
+        - VLLM_USAGE_SOURCE
+        - HF_TOKEN
+
+  #- block: "Run H100 Benchmark"
+    #key: block-h100
+    #depends_on: ~
+
+  - label: "H100"
+    # skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
+    agents:
+      queue: H100
+    depends_on: wait-for-container-image
+    if: build.branch == "main"
+    plugins:
+    - docker#v5.12.0:
+        image: public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT
+        command:
+        - bash
+        - .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+        mount-buildkite-agent: true
+        propagate-environment: true
+        ipc: host
+        gpus: all # see CUDA_VISIBLE_DEVICES for actual GPUs used
+        volumes:
+          - /data/benchmark-hf-cache:/root/.cache/huggingface
+        environment:
+        - VLLM_USAGE_SOURCE
+        - HF_TOKEN
+
+  # Premerge benchmark
+  - label: "A100"
+    # skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
+    agents:
+      queue: A100
+    depends_on: wait-for-container-image
+    if: build.branch != "main"
+    plugins:
+    - kubernetes:
+        podSpec:
+          priorityClassName: perf-benchmark
+          containers:
+          - image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
+            command:
+            - bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+            resources:
+              limits:
+                nvidia.com/gpu: 8
+            volumeMounts:
+            - name: devshm
+              mountPath: /dev/shm
+            env:
+            - name: VLLM_USAGE_SOURCE
+              value: ci-test
+            - name: HF_TOKEN
+              valueFrom:
+                secretKeyRef:
+                  name: hf-token-secret
+                  key: token
+          nodeSelector:
+            nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
+          volumes:
+          - name: devshm
+            emptyDir:
+              medium: Memory
+
+  - label: "H200"
+    # skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
+    agents:
+      queue: H200
+    depends_on: wait-for-container-image
+    if: build.branch != "main"
+    plugins:
+    - docker#v5.12.0:
+        image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
+        command:
+        - bash
+        - .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+        mount-buildkite-agent: true
+        propagate-environment: true
+        ipc: host
+        gpus: 4,5,6,7
+        volumes:
+          - /data/benchmark-hf-cache:/root/.cache/huggingface
+        environment:
+        - VLLM_USAGE_SOURCE
+        - HF_TOKEN
+
+  #- block: "Run H100 Benchmark"
+    #key: block-h100
+    #depends_on: ~
+
+  - label: "H100"
+    # skip: "use this flag to conditionally skip the benchmark step, useful for PR testing"
+    agents:
+      queue: H100
+    depends_on: wait-for-container-image
+    if: build.branch != "main"
+    plugins:
+    - docker#v5.12.0:
+        image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT
+        command:
+        - bash
+        - .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+        mount-buildkite-agent: true
+        propagate-environment: true
+        ipc: host
+        gpus: all # see CUDA_VISIBLE_DEVICES for actual GPUs used
+        volumes:
+          - /data/benchmark-hf-cache:/root/.cache/huggingface
+        environment:
+        - VLLM_USAGE_SOURCE
+        - HF_TOKEN

.buildkite/nightly-benchmarks/nightly-annotation.md

@@ -0,0 +1,28 @@
+# Nightly benchmark annotation
+
+## Description
+
+This file contains the downloading link for benchmarking results.
+
+- [benchmarking pipeline](artifact://nightly-pipeline.yaml)
+- [benchmarking results](artifact://results.zip)
+- [benchmarking code](artifact://nightly-benchmarks.zip)
+
+Please download the visualization scripts in the post
+
+## Results reproduction
+
+- Find the docker we use in `benchmarking pipeline`
+- Deploy the docker, and inside the docker:
+    - Download `nightly-benchmarks.zip`.
+    - In the same folder, run the following code:
+
+    ```bash
+    export HF_TOKEN=<your HF token>
+    apt update
+    apt install -y git
+    unzip nightly-benchmarks.zip
+    VLLM_SOURCE_CODE_LOC=./ bash .buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
+    ```
+
+And the results will be inside `./benchmarks/results`.

.buildkite/nightly-benchmarks/nightly-descriptions.md

@@ -0,0 +1,39 @@
+
+# Nightly benchmark
+
+This benchmark aims to:
+
+- Provide performance clarity: Provide clarity on which one (vllm, tensorrt-llm, lmdeploy and SGLang) leads in performance in what workload.
+- Be reproducible: one can run the exact same set of benchmarking commands inside the exact same docker by following reproducing instructions.
+
+Latest results: [results link](https://blog.vllm.ai/2024/09/05/perf-update.html), scroll to the end.
+
+Latest reproduction guilde: [github issue link](https://github.com/vllm-project/vllm/issues/8176)
+
+## Setup
+
+- Docker images:
+    - vLLM: `vllm/vllm-openai:v0.6.2`
+    - SGLang: `lmsysorg/sglang:v0.3.2-cu121`
+    - LMDeploy: `openmmlab/lmdeploy:v0.6.1-cu12`
+    - TensorRT-LLM: `nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3`
+        - *NOTE: we use r24.07 as the current implementation only works for this version. We are going to bump this up.*
+    - Check [nightly-pipeline.yaml](nightly-pipeline.yaml) for the concrete docker images, specs and commands we use for the benchmark.
+- Hardware
+    - 8x Nvidia A100 GPUs
+- Workload:
+    - Dataset
+        - ShareGPT dataset
+        - Prefill-heavy dataset (in average 462 input tokens, 16 tokens as output)
+        - Decode-heavy dataset (in average 462 input tokens, 256 output tokens)
+        - Check [nightly-tests.json](tests/nightly-tests.json) for the concrete configuration of datasets we use.
+    - Models: llama-3 8B, llama-3 70B.
+        - We do not use llama 3.1 as it is incompatible with trt-llm r24.07. ([issue](https://github.com/NVIDIA/TensorRT-LLM/issues/2105)).
+    - Average QPS (query per second): 2, 4, 8, 16, 32 and inf.
+        - Queries are randomly sampled, and arrival patterns are determined via Poisson process, but all with fixed random seed.
+    - Evaluation metrics: Throughput (higher the better), TTFT (time to the first token, lower the better), ITL (inter-token latency, lower the better).
+
+## Known issues
+
+- TRT-LLM crashes with Llama 3.1 8B [issue](https://github.com/NVIDIA/TensorRT-LLM/issues/2105).
+- TGI does not support `ignore-eos` flag.

.buildkite/nightly-benchmarks/nightly-pipeline.yaml

@@ -0,0 +1,196 @@
+common_pod_spec: &common_pod_spec
+  priorityClassName: perf-benchmark
+  nodeSelector:
+    nvidia.com/gpu.product: NVIDIA-A100-SXM4-80GB
+  volumes:
+    - name: devshm
+      emptyDir:
+        medium: Memory
+    - name: hf-cache
+      hostPath:
+        path: /root/.cache/huggingface
+        type: Directory
+
+common_container_settings: &common_container_settings
+  command:
+    - bash .buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
+  resources:
+    limits:
+      nvidia.com/gpu: 8
+  volumeMounts:
+    - name: devshm
+      mountPath: /dev/shm
+    - name: hf-cache
+      mountPath: /root/.cache/huggingface
+  env:
+    - name: VLLM_USAGE_SOURCE
+      value: ci-test
+    - name: HF_HOME
+      value: /root/.cache/huggingface
+    - name: VLLM_SOURCE_CODE_LOC
+      value: /workspace/build/buildkite/vllm/performance-benchmark
+    - name: HF_TOKEN
+      valueFrom:
+        secretKeyRef:
+          name: hf-token-secret
+          key: token
+
+steps:
+  - block: ":rocket: Ready for comparing vllm against alternatives? This will take 4 hours."
+
+
+
+  - label: "A100 vllm step 10"
+    priority: 100
+    agents:
+      queue: A100
+    plugins:
+      - kubernetes:
+          podSpec:
+            <<: *common_pod_spec
+            containers:
+              - image: vllm/vllm-openai:v0.6.2
+                <<: *common_container_settings
+
+
+
+  - label: "A100 sglang benchmark"
+    priority: 100
+    agents:
+      queue: A100
+    plugins:
+      - kubernetes:
+          podSpec:
+            <<: *common_pod_spec
+            containers:
+              - image: lmsysorg/sglang:v0.3.2-cu121
+                <<: *common_container_settings
+
+  - label: "A100 lmdeploy benchmark"
+    priority: 100
+    agents:
+      queue: A100
+    plugins:
+      - kubernetes:
+          podSpec:
+            <<: *common_pod_spec
+            containers:
+              - image: openmmlab/lmdeploy:v0.6.1-cu12
+                <<: *common_container_settings
+
+
+
+
+  - label: "A100 trt llama-8B"
+    priority: 100
+    agents:
+      queue: A100
+    plugins:
+      - kubernetes:
+          podSpec:
+            <<: *common_pod_spec
+            containers:
+              - image: nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3
+                <<: *common_container_settings
+                env:
+                  - name: VLLM_USAGE_SOURCE
+                    value: ci-test
+                  - name: HF_HOME
+                    value: /root/.cache/huggingface
+                  - name: VLLM_SOURCE_CODE_LOC
+                    value: /workspace/build/buildkite/vllm/performance-benchmark
+                  - name: HF_TOKEN
+                    valueFrom:
+                      secretKeyRef:
+                        name: hf-token-secret
+                        key: token
+                  - name: TEST_SELECTOR
+                    value: "llama8B"
+
+
+  - label: "A100 trt llama-70B"
+    priority: 100
+    agents:
+      queue: A100
+    plugins:
+      - kubernetes:
+          podSpec:
+            <<: *common_pod_spec
+            containers:
+              - image: nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3
+                <<: *common_container_settings
+                env:
+                  - name: VLLM_USAGE_SOURCE
+                    value: ci-test
+                  - name: HF_HOME
+                    value: /root/.cache/huggingface
+                  - name: VLLM_SOURCE_CODE_LOC
+                    value: /workspace/build/buildkite/vllm/performance-benchmark
+                  - name: HF_TOKEN
+                    valueFrom:
+                      secretKeyRef:
+                        name: hf-token-secret
+                        key: token
+                  - name: TEST_SELECTOR
+                    value: "llama70B"
+
+
+  # FIXME(Kuntai): uncomment this after NVIDIA gives us their test docker image 
+  # - label: "A100 trt benchmark"
+  #   priority: 100
+  #   agents:
+  #     queue: A100
+  #   plugins:
+  #     - kubernetes:
+  #         podSpec:
+  #           <<: *common_pod_spec
+  #           containers:
+  #             - image: nvcr.io/nvidia/tritonserver:24.07-trtllm-python-py3
+  #               <<: *common_container_settings
+
+
+  # FIXME(Kuntai): uncomment this after TGI supports `--ignore-eos`.
+  # - label: "A100 tgi benchmark"
+  #   priority: 100
+  #   agents:
+  #     queue: A100
+  #   plugins:
+  #     - kubernetes:
+  #         podSpec:
+  #           <<: *common_pod_spec
+  #           containers:
+  #             - image: ghcr.io/huggingface/text-generation-inference:2.2.0
+  #               <<: *common_container_settings
+        
+  - wait
+
+  - label: "Collect the results"
+    priority: 100
+    agents:
+      queue: A100
+    plugins:
+      - kubernetes:
+          podSpec:
+            <<: *common_pod_spec
+            containers:
+            - image: vllm/vllm-openai:v0.5.0.post1
+              command:
+              - bash .buildkite/nightly-benchmarks/scripts/nightly-annotate.sh
+              resources:
+                limits:
+                  nvidia.com/gpu: 8
+              volumeMounts:
+              - name: devshm
+                mountPath: /dev/shm
+              env:
+              - name: VLLM_USAGE_SOURCE
+                value: ci-test
+              - name: VLLM_SOURCE_CODE_LOC
+                value: /workspace/build/buildkite/vllm/performance-benchmark
+              - name: HF_TOKEN
+                valueFrom:
+                  secretKeyRef:
+                    name: hf-token-secret
+                    key: token
+
+  - block: ":rocket: check the results!"

.buildkite/nightly-benchmarks/scripts/compare-json-results.py

@@ -7,7 +7,6 @@ from importlib import util
 
 import pandas as pd
 
-pd.options.display.float_format = "{:.2f}".format
 plotly_found = util.find_spec("plotly.express") is not None
 
 
@@ -110,10 +109,7 @@ def compare_data_columns(
         if len(compare_frames) >= 2:
             base = compare_frames[0]
             current = compare_frames[-1]
-            if "P99" in data_column or "Median" in data_column:
-                ratio = base / current  # for latency
-            else:
-                ratio = current / base
+            ratio = current / base
             ratio = ratio.mask(base == 0)  # avoid inf when baseline is 0
             ratio.name = f"Ratio 1 vs {len(compare_frames)}"
             frames.append(ratio)
@@ -203,71 +199,6 @@ def split_json_by_tp_pp(
     return saved_paths
 
 
-def _add_limit_line(fig, y_value, label):
-    # Visible dashed line + annotation
-    fig.add_hline(
-        y=y_value,
-        line_dash="dash",
-        line_color="red" if "ttft" in label.lower() else "blue",
-        annotation_text=f"{label}: {y_value} ms",
-        annotation_position="top left",
-    )
-    # Optional: add a legend item (as a transparent helper trace)
-    if plot and plotly_found:
-        import plotly.graph_objects as go
-
-        fig.add_trace(
-            go.Scatter(
-                x=[None],
-                y=[None],
-                mode="lines",
-                line=dict(
-                    dash="dash", color="red" if "ttft" in label.lower() else "blue"
-                ),
-                name=f"{label}",
-            )
-        )
-
-
-def _find_concurrency_col(df: pd.DataFrame) -> str:
-    for c in [
-        "# of max concurrency.",
-        "# of max concurrency",
-        "Max Concurrency",
-        "max_concurrency",
-        "Concurrency",
-    ]:
-        if c in df.columns:
-            return c
-    # Fallback: guess an integer-like column (harmless if unused)
-    for c in df.columns:
-        if df[c].dtype.kind in "iu" and df[c].nunique() > 1 and df[c].min() >= 1:
-            return c
-    return "# of max concurrency."
-
-
-def _highlight_threshold(
-    df: pd.DataFrame, threshold: float
-) -> "pd.io.formats.style.Styler":
-    """Highlight numeric per-configuration columns with value <= threshold."""
-    conc_col = _find_concurrency_col(df)
-    key_cols = [
-        c
-        for c in ["Model", "Dataset Name", "Input Len", "Output Len", conc_col]
-        if c in df.columns
-    ]
-    conf_cols = [
-        c for c in df.columns if c not in key_cols and not str(c).startswith("Ratio")
-    ]
-    conf_cols = [c for c in conf_cols if pd.api.types.is_numeric_dtype(df[c])]
-    return df.style.map(
-        lambda v: "background-color:#e6ffe6;font-weight:bold;"
-        if pd.notna(v) and v <= threshold
-        else "",
-        subset=conf_cols,
-    )
-
-
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     parser.add_argument(
@@ -289,26 +220,6 @@ if __name__ == "__main__":
         default="# of max concurrency.",
         help="column name to use as X Axis in comparison graph",
     )
-    parser.add_argument(
-        "-l",
-        "--latency",
-        type=str,
-        default="p99",
-        help="take median|p99 for latency like TTFT/TPOT",
-    )
-    parser.add_argument(
-        "--ttft-max-ms",
-        type=float,
-        default=3000.0,
-        help="Reference limit for TTFT plots (ms)",
-    )
-    parser.add_argument(
-        "--tpot-max-ms",
-        type=float,
-        default=100.0,
-        help="Reference limit for TPOT plots (ms)",
-    )
-
     args = parser.parse_args()
 
     drop_column = "P99"
@@ -323,22 +234,12 @@ if __name__ == "__main__":
         "# of max concurrency.",
         "qps",
     ]
-
-    if "median" in args.latency:
-        data_cols_to_compare = ["Output Tput (tok/s)", "Median TTFT (ms)", "Median"]
-        html_msgs_for_data_cols = [
-            "Compare Output Tokens /n",
-            "Median TTFT /n",
-            "Median TPOT /n",
-        ]
-        drop_column = "P99"
-    elif "p99" in args.latency:
-        data_cols_to_compare = ["Output Tput (tok/s)", "P99 TTFT (ms)", "P99"]
-        html_msgs_for_data_cols = [
-            "Compare Output Tokens /n",
-            "P99 TTFT /n",
-            "P99 TPOT /n",
-        ]
+    data_cols_to_compare = ["Output Tput (tok/s)", "Median TTFT (ms)", "Median"]
+    html_msgs_for_data_cols = [
+        "Compare Output Tokens /n",
+        "Median TTFT /n",
+        "Median TPOT /n",
+    ]
 
     if len(args.file) == 1:
         files = split_json_by_tp_pp(args.file[0], output_root="splits")
@@ -374,83 +275,33 @@ if __name__ == "__main__":
                     f"Expected subset: {filtered_info_cols}, "
                     f"but DataFrame has: {list(output_df.columns)}"
                 )
-            # output_df_sorted = output_df.sort_values(by=existing_group_cols)
-            output_df_sorted = output_df.sort_values(by=args.xaxis)
+            output_df_sorted = output_df.sort_values(by=existing_group_cols)
             output_groups = output_df_sorted.groupby(existing_group_cols, dropna=False)
             for name, group in output_groups:
-                group_name = (
-                    ",".join(map(str, name)).replace(",", "_").replace("/", "-")
-                )
-                group_html_name = "perf_comparison_" + group_name + ".html"
-
-                metric_name = str(data_cols_to_compare[i]).lower()
-                if "tok/s" in metric_name:
-                    html = group.to_html()
-                elif "ttft" in metric_name:
-                    styler = _highlight_threshold(group, args.ttft_max_ms).format(
-                        {c: "{:.2f}" for c in group.select_dtypes("number").columns},
-                        na_rep="—",
-                    )
-                    html = styler.to_html(
-                        table_attributes='border="1" class="dataframe"'
-                    )
-                elif (
-                    "tpot" in metric_name
-                    or "median" in metric_name
-                    or "p99" in metric_name
-                ):
-                    styler = _highlight_threshold(group, args.tpot_max_ms).format(
-                        {c: "{:.2f}" for c in group.select_dtypes("number").columns},
-                        na_rep="—",
-                    )
-                    html = styler.to_html(
-                        table_attributes='border="1" class="dataframe"'
-                    )
-
+                html = group.to_html()
                 text_file.write(html_msgs_for_data_cols[i])
                 text_file.write(html)
-                with open(group_html_name, "a+") as sub_text_file:
-                    sub_text_file.write(html_msgs_for_data_cols[i])
-                    sub_text_file.write(html)
 
-                    if plot and plotly_found:
-                        import plotly.express as px
+                if plot and plotly_found:
+                    import plotly.express as px
 
-                        df = group[raw_data_cols]
-                        df_sorted = df.sort_values(by=info_cols[y_axis_index])
-                        # Melt DataFrame for plotting
-                        df_melted = df_sorted.melt(
-                            id_vars=info_cols[y_axis_index],
-                            var_name="Configuration",
-                            value_name=data_cols_to_compare[i],
-                        )
-                        title = (
-                            data_cols_to_compare[i] + " vs " + info_cols[y_axis_index]
-                        )
-                        # Create Plotly line chart
-                        fig = px.line(
-                            df_melted,
-                            x=info_cols[y_axis_index],
-                            y=data_cols_to_compare[i],
-                            color="Configuration",
-                            title=title,
-                            markers=True,
-                        )
-
-                        # ---- Add threshold lines based on metric name ----
-                        if "ttft" in metric_name:
-                            _add_limit_line(fig, args.ttft_max_ms, "TTFT limit")
-                        elif (
-                            "tpot" in metric_name
-                            or "median" in metric_name
-                            or "p99" in metric_name
-                        ):
-                            _add_limit_line(fig, args.tpot_max_ms, "TPOT limit")
-
-                        # Export to HTML
-                        text_file.write(
-                            fig.to_html(full_html=True, include_plotlyjs="cdn")
-                        )
-                        sub_text_file.write(
-                            fig.to_html(full_html=True, include_plotlyjs="cdn")
-                        )
+                    df = group[raw_data_cols]
+                    df_sorted = df.sort_values(by=info_cols[y_axis_index])
+                    # Melt DataFrame for plotting
+                    df_melted = df_sorted.melt(
+                        id_vars=info_cols[y_axis_index],
+                        var_name="Configuration",
+                        value_name=data_cols_to_compare[i],
+                    )
+                    title = data_cols_to_compare[i] + " vs " + info_cols[y_axis_index]
+                    # Create Plotly line chart
+                    fig = px.line(
+                        df_melted,
+                        x=info_cols[y_axis_index],
+                        y=data_cols_to_compare[i],
+                        color="Configuration",
+                        title=title,
+                        markers=True,
+                    )
+                    # Export to HTML
+                    text_file.write(fig.to_html(full_html=True, include_plotlyjs="cdn"))

.buildkite/nightly-benchmarks/scripts/convert-results-json-to-markdown.py

@@ -63,11 +63,9 @@ serving_column_mapping = {
     "mean_ttft_ms": "Mean TTFT (ms)",
     "median_ttft_ms": "Median TTFT (ms)",
     "p99_ttft_ms": "P99 TTFT (ms)",
-    "std_ttft_ms": "STD TTFT (ms)",
     "mean_tpot_ms": "Mean TPOT (ms)",
     "median_tpot_ms": "Median",
     "p99_tpot_ms": "P99",
-    "std_tpot_ms": "STD TPOT (ms)",
     "mean_itl_ms": "Mean ITL (ms)",
     "median_itl_ms": "Median ITL (ms)",
     "p99_itl_ms": "P99 ITL (ms)",
@@ -370,7 +368,7 @@ if __name__ == "__main__":
         # The GPUs sometimes come in format of "GPUTYPE\nGPUTYPE\n...",
         # we want to turn it into "8xGPUTYPE"
         df["GPU"] = df["GPU"].apply(
-            lambda x: "{}x{}".format(len(x.split("\n")), x.split("\n")[0])
+            lambda x: f"{len(x.split('\n'))}x{x.split('\n')[0]}"
         )
 
     # get markdown tables
@@ -392,7 +390,7 @@ if __name__ == "__main__":
     json_file = "benchmark_results.json"
     with open(results_folder / md_file, "w") as f:
         results = read_markdown(
-            "../.buildkite/performance-benchmarks/"
+            "../.buildkite/nightly-benchmarks/"
             + "performance-benchmarks-descriptions.md"
         )
         results = results.format(

.buildkite/nightly-benchmarks/scripts/download-tokenizer.py

@@ -0,0 +1,26 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import argparse
+
+from transformers import AutoTokenizer
+
+
+def main(model, cachedir):
+    # Load the tokenizer and save it to the specified directory
+    tokenizer = AutoTokenizer.from_pretrained(model)
+    tokenizer.save_pretrained(cachedir)
+    print(f"Tokenizer saved to {cachedir}")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(
+        description="Download and save Hugging Face tokenizer"
+    )
+    parser.add_argument("--model", type=str, required=True, help="Name of the model")
+    parser.add_argument(
+        "--cachedir", type=str, required=True, help="Directory to save the tokenizer"
+    )
+
+    args = parser.parse_args()
+    main(args.model, args.cachedir)

.buildkite/nightly-benchmarks/scripts/generate-nightly-markdown.py

@@ -0,0 +1,97 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import argparse
+import json
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+from tabulate import tabulate
+
+
+def parse_arguments():
+    parser = argparse.ArgumentParser(
+        description="Parse command line arguments for summary-nightly-results script."
+    )
+    parser.add_argument(
+        "--results-folder",
+        type=str,
+        required=True,
+        help="The folder where the results are stored.",
+    )
+    parser.add_argument(
+        "--description", type=str, required=True, help="Description of the results."
+    )
+
+    args = parser.parse_args()
+    return args
+
+
+def get_perf(df, method, model, metric):
+    means = []
+
+    for qps in [2, 4, 8, 16, "inf"]:
+        target = df["Test name"].str.contains(model)
+        target = target & df["Engine"].str.contains(method)
+        target = target & df["Test name"].str.contains("qps_" + str(qps))
+        filtered_df = df[target]
+
+        if filtered_df.empty:
+            means.append(0.0)
+        else:
+            means.append(filtered_df[metric].values[0])
+
+    return np.array(means)
+
+
+def get_perf_w_std(df, method, model, metric):
+    if metric in ["TTFT", "ITL"]:
+        mean = get_perf(df, method, model, "Mean " + metric + " (ms)")
+        mean = mean.tolist()
+        std = get_perf(df, method, model, "Std " + metric + " (ms)")
+        if std.mean() == 0:
+            std = None
+        success = get_perf(df, method, model, "Successful req.")
+        if std is not None:
+            std = std / np.sqrt(success)
+            std = std.tolist()
+
+    else:
+        assert metric == "Tput"
+        mean = get_perf(df, method, model, "Input Tput (tok/s)") + get_perf(
+            df, method, model, "Output Tput (tok/s)"
+        )
+        mean = mean.tolist()
+        std = None
+
+    return mean, std
+
+
+def main(args):
+    results_folder = Path(args.results_folder)
+
+    results = []
+
+    # collect results
+    for test_file in results_folder.glob("*_nightly_results.json"):
+        with open(test_file) as f:
+            results = results + json.loads(f.read())
+
+    # generate markdown table
+    df = pd.DataFrame.from_dict(results)
+
+    md_table = tabulate(df, headers="keys", tablefmt="pipe", showindex=False)
+
+    with open(args.description) as f:
+        description = f.read()
+
+    description = description.format(nightly_results_benchmarking_table=md_table)
+
+    with open("nightly_results.md", "w") as f:
+        f.write(description)
+
+
+if __name__ == "__main__":
+    args = parse_arguments()
+    main(args)

.buildkite/nightly-benchmarks/scripts/get-lmdeploy-modelname.py

@@ -0,0 +1,9 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from lmdeploy.serve.openai.api_client import APIClient
+
+api_client = APIClient("http://localhost:8000")
+model_name = api_client.available_models[0]
+
+print(model_name)

.buildkite/nightly-benchmarks/scripts/launch-server.sh

@@ -181,14 +181,18 @@ launch_vllm_server() {
   if echo "$common_params" | jq -e 'has("fp8")' >/dev/null; then
     echo "Key 'fp8' exists in common params. Use neuralmagic fp8 model for convenience."
     model=$(echo "$common_params" | jq -r '.neuralmagic_quantized_model')
-    server_command="vllm serve $model \
+    server_command="python3 \
+        -m vllm.entrypoints.openai.api_server \
         -tp $tp \
+        --model $model \
         --port $port \
         $server_args"
   else
     echo "Key 'fp8' does not exist in common params."
-    server_command="vllm serve $model \
+    server_command="python3 \
+        -m vllm.entrypoints.openai.api_server \
         -tp $tp \
+        --model $model \
         --port $port \
         $server_args"
   fi

.buildkite/nightly-benchmarks/scripts/nightly-annotate.sh

@@ -0,0 +1,78 @@
+#!/bin/bash
+
+set -ex
+set -o pipefail
+
+
+main() {
+
+    (which wget && which curl) || (apt-get update && apt-get install -y wget curl)
+    (which jq) || (apt-get update && apt-get -y install jq)
+    (which zip) || (apt-get install -y zip)
+
+    if [ ! -f /workspace/buildkite-agent ]; then
+        echo "buildkite-agent binary not found. Skip plotting the results."
+        exit 0
+    fi
+
+    # initial annotation
+    #description="$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/nightly-descriptions.md"
+
+    # download results
+    cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
+    mkdir -p results/
+    /workspace/buildkite-agent artifact download 'results/*nightly_results.json' results/
+    ls
+    ls results/
+
+    # upload benchmark results
+    zip -r results.zip results/
+    /workspace/buildkite-agent artifact upload "results.zip"
+
+    # upload benchmarking scripts
+    cd "$VLLM_SOURCE_CODE_LOC/"
+    zip -r nightly-benchmarks.zip .buildkite/ benchmarks/
+    /workspace/buildkite-agent artifact upload "nightly-benchmarks.zip"
+
+    cd "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/"
+    # upload benchmarking pipeline
+    /workspace/buildkite-agent artifact upload "nightly-pipeline.yaml"
+
+    cd "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/"
+    /workspace/buildkite-agent annotate --style "success" --context "nightly-benchmarks-results" --append < nightly-annotation.md
+    
+
+
+    # The figures should be generated by a separate process outside the CI/CD pipeline
+
+    # # generate figures
+    # python3 -m pip install tabulate pandas matplotlib
+
+    # python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/generate-nightly-markdown.py \
+    #     --description $description \
+    #     --results-folder results/ 
+
+
+    # python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/plot-nightly-results.py \
+    #     --description $description \
+    #     --results-folder results/ \
+    #     --dataset sharegpt
+
+    # python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/plot-nightly-results.py \
+    #     --description $description \
+    #     --results-folder results/ \
+    #     --dataset sonnet_2048_128
+
+    # python3 $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/plot-nightly-results.py \
+    #     --description $description \
+    #     --results-folder results/ \
+    #     --dataset sonnet_128_2048
+    
+    # # upload results and figures
+    # /workspace/buildkite-agent artifact upload "nightly_results*.png"
+    # /workspace/buildkite-agent artifact upload $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/nightly-pipeline.yaml
+    # /workspace/buildkite-agent artifact upload $VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/tests/nightly-tests.json
+    # /workspace/buildkite-agent annotate --style "success" --context "nightly-benchmarks-results" --append < nightly_results.md
+}
+
+main "$@"

.buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh

@@ -0,0 +1,464 @@
+#!/bin/bash
+
+set -o pipefail
+set -x
+
+check_gpus() {
+  # check the number of GPUs and GPU type.
+  declare -g gpu_count=$(nvidia-smi --list-gpus | wc -l)
+  if [[ $gpu_count -gt 0 ]]; then
+    echo "GPU found."
+  else
+    echo "Need at least 1 GPU to run benchmarking."
+    exit 1
+  fi
+  declare -g gpu_type="$(nvidia-smi --query-gpu=name --format=csv,noheader | awk '{print $2}')"
+  echo "GPU type is $gpu_type"
+}
+
+check_hf_token() {
+  # check if HF_TOKEN is available and valid
+  if [[ -z "$HF_TOKEN" ]]; then
+    echo "Error: HF_TOKEN is not set."
+    exit 1
+  elif [[ ! "$HF_TOKEN" =~ ^hf_ ]]; then
+    echo "Error: HF_TOKEN does not start with 'hf_'."
+    exit 1
+  else
+    echo "HF_TOKEN is set and valid."
+  fi
+}
+
+
+upload_to_buildkite() {
+  # upload the benchmarking results to buildkite
+
+  # if the agent binary is not found, skip uploading the results, exit 0
+  if [ ! -f /workspace/buildkite-agent ]; then
+    echo "buildkite-agent binary not found. Skip uploading the results."
+    return 0
+  fi
+  # /workspace/buildkite-agent annotate --style "success" --context "benchmark-results" --append < $RESULTS_FOLDER/${CURRENT_LLM_SERVING_ENGINE}_nightly_results.md
+  /workspace/buildkite-agent artifact upload "$RESULTS_FOLDER/*"
+}
+
+
+get_current_llm_serving_engine() {
+
+  if which lmdeploy >/dev/null; then
+    echo "Container: lmdeploy"
+    export CURRENT_LLM_SERVING_ENGINE=lmdeploy
+    return
+  fi
+
+  if [ -e /tgi-entrypoint.sh ]; then
+    echo "Container: tgi"
+    export CURRENT_LLM_SERVING_ENGINE=tgi
+    return
+  fi
+
+  if which trtllm-build >/dev/null; then
+    echo "Container: tensorrt-llm"
+    export CURRENT_LLM_SERVING_ENGINE=trt
+    return
+  fi
+
+  if [ -e /sgl-workspace ]; then
+    echo "Container: sglang"
+    export CURRENT_LLM_SERVING_ENGINE=sglang
+    return
+  fi
+
+  if [ -e /vllm-workspace ]; then
+    echo "Container: vllm"
+    # move to a completely irrelevant directory, to avoid import vllm from current folder
+    export CURRENT_LLM_SERVING_ENGINE=vllm
+
+    return
+  fi
+}
+
+json2args() {
+  # transforms the JSON string to command line args, and '_' is replaced to '-'
+  # example:
+  # input: { "model": "meta-llama/Llama-2-7b-chat-hf", "tensor_parallel_size": 1 }
+  # output: --model meta-llama/Llama-2-7b-chat-hf --tensor-parallel-size 1
+  local json_string=$1
+  local args=$(
+    echo "$json_string" | jq -r '
+      to_entries |
+      map("--" + (.key | gsub("_"; "-")) + " " + (.value | tostring)) |
+      join(" ")
+    '
+  )
+  echo "$args"
+}
+
+kill_gpu_processes() {
+  pkill -f '[p]ython'
+  pkill -f '[p]ython3'
+  pkill -f '[t]ritonserver'
+  pkill -f '[p]t_main_thread'
+  pkill -f '[t]ext-generation'
+  pkill -f '[l]mdeploy'
+  # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
+  pkill -f '[V]LLM'
+
+  while [ "$(nvidia-smi --query-gpu=memory.used --format=csv,noheader,nounits | head -n 1)" -ge 1000 ]; do
+    sleep 1
+  done
+}
+
+wait_for_server() {
+  # wait for vllm server to start
+  # return 1 if vllm server crashes
+  timeout 1200 bash -c '
+    until curl -s localhost:8000/v1/completions > /dev/null; do
+      sleep 1
+    done' && return 0 || return 1
+}
+
+ensure_installed() {
+  # Ensure that the given command is installed by apt-get
+  local cmd=$1
+  if ! which "$cmd" >/dev/null; then
+    apt-get update && apt-get install -y "$cmd"
+  fi
+}
+
+run_serving_tests() {
+  # run serving tests using `vllm bench serve` command
+  # $1: a json file specifying serving test cases
+
+  local serving_test_file
+  serving_test_file=$1
+
+  # Iterate over serving tests
+  jq -c '.[]' "$serving_test_file" | while read -r params; do
+    # get the test name, and append the GPU type back to it.
+    test_name=$(echo "$params" | jq -r '.test_name')
+
+    # if TEST_SELECTOR is set, only run the test cases that match the selector
+    if [[ -n "$TEST_SELECTOR" ]] && [[ ! "$test_name" =~ $TEST_SELECTOR ]]; then
+      echo "Skip test case $test_name."
+      continue
+    fi
+
+    # prepend the current serving engine to the test name
+    test_name=${CURRENT_LLM_SERVING_ENGINE}_${test_name}
+
+    # get common parameters
+    common_params=$(echo "$params" | jq -r '.common_parameters')
+    model=$(echo "$common_params" | jq -r '.model')
+    tp=$(echo "$common_params" | jq -r '.tp')
+    dataset_name=$(echo "$common_params" | jq -r '.dataset_name')
+    dataset_path=$(echo "$common_params" | jq -r '.dataset_path')
+    port=$(echo "$common_params" | jq -r '.port')
+    num_prompts=$(echo "$common_params" | jq -r '.num_prompts')
+    reuse_server=$(echo "$common_params" | jq -r '.reuse_server')
+
+    # get client and server arguments
+    server_params=$(echo "$params" | jq -r ".${CURRENT_LLM_SERVING_ENGINE}_server_parameters")
+    client_params=$(echo "$params" | jq -r ".${CURRENT_LLM_SERVING_ENGINE}_client_parameters")
+    client_args=$(json2args "$client_params")
+    qps_list=$(echo "$params" | jq -r '.qps_list')
+    qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
+    echo "Running over qps list $qps_list"
+
+    # check if there is enough GPU to run the test
+    if [[ $gpu_count -lt $tp ]]; then
+      echo "Required num-shard $tp but only $gpu_count GPU found. Skip testcase $test_name."
+      continue
+    fi
+
+    if [[ $reuse_server == "true" ]]; then
+      echo "Reuse previous server for test case $test_name"
+    else
+      kill_gpu_processes
+      bash "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/launch-server.sh" \
+        "$server_params" "$common_params"
+    fi
+
+    if wait_for_server; then
+      echo ""
+      echo "$CURRENT_LLM_SERVING_ENGINE server is up and running."
+    else
+      echo ""
+      echo "$CURRENT_LLM_SERVING_ENGINE failed to start within the timeout period."
+      break
+    fi
+
+    # prepare tokenizer
+    # this is required for lmdeploy.
+    cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
+    rm -rf /tokenizer_cache
+    mkdir /tokenizer_cache
+    python3 ../.buildkite/nightly-benchmarks/scripts/download-tokenizer.py \
+      --model "$model" \
+      --cachedir /tokenizer_cache
+    cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
+
+
+    # change model name for lmdeploy (it will not follow standard hf name)
+    if [[ "$CURRENT_LLM_SERVING_ENGINE" == "lmdeploy" ]]; then
+      model=$(python ../.buildkite/nightly-benchmarks/scripts/get-lmdeploy-modelname.py)
+    fi
+
+    # iterate over different QPS
+    for qps in $qps_list; do
+      # remove the surrounding single quote from qps
+      if [[ "$qps" == *"inf"* ]]; then
+        echo "qps was $qps"
+        qps="inf"
+        echo "now qps is $qps"
+      fi
+
+      new_test_name=$test_name"_qps_"$qps
+
+      backend=$CURRENT_LLM_SERVING_ENGINE
+
+      if [[ $backend = "trt" ]]; then
+        backend="tensorrt-llm"
+      fi
+
+      if [[ "$backend" == *"vllm"* ]]; then
+        backend="vllm"
+      fi
+
+      if [[ "$dataset_name" = "sharegpt" ]]; then
+
+        client_command="vllm bench serve \
+          --backend $backend \
+          --tokenizer /tokenizer_cache \
+          --model $model \
+          --dataset-name $dataset_name \
+          --dataset-path $dataset_path \
+          --num-prompts $num_prompts \
+          --port $port \
+          --save-result \
+          --result-dir $RESULTS_FOLDER \
+          --result-filename ${new_test_name}.json \
+          --request-rate $qps \
+          --ignore-eos \
+          $client_args"
+
+      elif [[ "$dataset_name" = "sonnet" ]]; then
+
+        sonnet_input_len=$(echo "$common_params" | jq -r '.sonnet_input_len')
+        sonnet_output_len=$(echo "$common_params" | jq -r '.sonnet_output_len')
+        sonnet_prefix_len=$(echo "$common_params" | jq -r '.sonnet_prefix_len')
+
+        client_command="vllm bench serve \
+          --backend $backend \
+          --tokenizer /tokenizer_cache \
+          --model $model \
+          --dataset-name $dataset_name \
+          --dataset-path $dataset_path \
+          --num-prompts $num_prompts \
+          --sonnet-input-len $sonnet_input_len \
+          --sonnet-output-len $sonnet_output_len \
+          --sonnet-prefix-len $sonnet_prefix_len \
+          --port $port \
+          --save-result \
+          --result-dir $RESULTS_FOLDER \
+          --result-filename ${new_test_name}.json \
+          --request-rate $qps \
+          --ignore-eos \
+          $client_args"
+
+      else
+
+        echo "The dataset name must be either 'sharegpt' or 'sonnet'. Got $dataset_name."
+        exit 1
+
+      fi
+
+
+
+      echo "Running test case $test_name with qps $qps"
+      echo "Client command: $client_command"
+
+      eval "$client_command"
+
+      server_command="None"
+
+      # record the benchmarking commands
+      jq_output=$(jq -n \
+        --arg server "$server_command" \
+        --arg client "$client_command" \
+        --arg gpu "$gpu_type" \
+        --arg engine "$CURRENT_LLM_SERVING_ENGINE" \
+        '{
+          server_command: $server,
+          client_command: $client,
+          gpu_type: $gpu,
+          engine: $engine
+        }')
+      echo "$jq_output" >"$RESULTS_FOLDER/${new_test_name}.commands"
+
+    done
+
+  done
+
+  kill_gpu_processes
+}
+
+run_genai_perf_tests() {
+  # run genai-perf tests
+
+  # $1: a json file specifying genai-perf test cases
+  local genai_perf_test_file
+  genai_perf_test_file=$1
+
+  # Iterate over genai-perf tests
+  jq -c '.[]' "$genai_perf_test_file" | while read -r params; do
+    # get the test name, and append the GPU type back to it.
+    test_name=$(echo "$params" | jq -r '.test_name')
+
+    # if TEST_SELECTOR is set, only run the test cases that match the selector
+    if [[ -n "$TEST_SELECTOR" ]] && [[ ! "$test_name" =~ $TEST_SELECTOR ]]; then
+      echo "Skip test case $test_name."
+      continue
+    fi
+
+    # prepend the current serving engine to the test name
+    test_name=${CURRENT_LLM_SERVING_ENGINE}_${test_name}
+
+    # get common parameters
+    common_params=$(echo "$params" | jq -r '.common_parameters')
+    model=$(echo "$common_params" | jq -r '.model')
+    tp=$(echo "$common_params" | jq -r '.tp')
+    dataset_name=$(echo "$common_params" | jq -r '.dataset_name')
+    dataset_path=$(echo "$common_params" | jq -r '.dataset_path')
+    port=$(echo "$common_params" | jq -r '.port')
+    num_prompts=$(echo "$common_params" | jq -r '.num_prompts')
+    reuse_server=$(echo "$common_params" | jq -r '.reuse_server')
+
+    # get client and server arguments
+    server_params=$(echo "$params" | jq -r ".${CURRENT_LLM_SERVING_ENGINE}_server_parameters")
+    qps_list=$(echo "$params" | jq -r '.qps_list')
+    qps_list=$(echo "$qps_list" | jq -r '.[] | @sh')
+    echo "Running over qps list $qps_list"
+
+    # check if there is enough GPU to run the test
+    if [[ $gpu_count -lt $tp ]]; then
+      echo "Required num-shard $tp but only $gpu_count GPU found. Skip testcase $test_name."
+      continue
+    fi
+
+    if [[ $reuse_server == "true" ]]; then
+      echo "Reuse previous server for test case $test_name"
+    else
+      kill_gpu_processes
+      bash "$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/scripts/launch-server.sh" \
+        "$server_params" "$common_params"
+    fi
+
+    if wait_for_server; then
+      echo ""
+      echo "$CURRENT_LLM_SERVING_ENGINE server is up and running."
+    else
+      echo ""
+      echo "$CURRENT_LLM_SERVING_ENGINE failed to start within the timeout period."
+      break
+    fi
+
+    # iterate over different QPS
+    for qps in $qps_list; do
+      # remove the surrounding single quote from qps
+      if [[ "$qps" == *"inf"* ]]; then
+        echo "qps was $qps"
+        qps=$num_prompts
+        echo "now qps is $qps"
+      fi
+
+      new_test_name=$test_name"_qps_"$qps
+      backend=$CURRENT_LLM_SERVING_ENGINE
+
+      if [[ "$backend" == *"vllm"* ]]; then
+        backend="vllm"
+      fi
+      #TODO: add output dir.
+      client_command="genai-perf profile \
+        -m $model \
+        --service-kind openai \
+        --backend "$backend" \
+        --endpoint-type chat \
+        --streaming \
+        --url localhost:$port \
+        --request-rate $qps \
+        --num-prompts $num_prompts \
+      "
+
+    echo "Client command: $client_command"
+
+    eval "$client_command"
+
+    #TODO: process/record outputs
+    done
+  done
+
+  kill_gpu_processes
+
+}
+
+prepare_dataset() {
+
+  # download sharegpt dataset
+  cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
+  wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
+
+  # duplicate sonnet by 4x, to allow benchmarking with input length 2048
+  cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
+  echo "" > sonnet_4x.txt
+  for _ in {1..4}
+  do
+    cat sonnet.txt >> sonnet_4x.txt
+  done
+
+}
+
+main() {
+
+  # check if the environment variable is successfully injected from yaml
+
+  check_gpus
+  check_hf_token
+  get_current_llm_serving_engine
+
+  pip install -U transformers
+
+  pip install -r requirements/dev.txt
+  which genai-perf
+
+  # check storage
+  df -h
+
+  ensure_installed wget
+  ensure_installed curl
+  ensure_installed jq
+  # genai-perf dependency
+  ensure_installed libb64-0d
+
+  prepare_dataset
+
+  cd "$VLLM_SOURCE_CODE_LOC/benchmarks"
+  declare -g RESULTS_FOLDER=results/
+  mkdir -p $RESULTS_FOLDER
+  BENCHMARK_ROOT="$VLLM_SOURCE_CODE_LOC/.buildkite/nightly-benchmarks/"
+
+  # run the test
+  run_serving_tests "$BENCHMARK_ROOT/tests/nightly-tests.json"
+
+  # run genai-perf tests
+  run_genai_perf_tests "$BENCHMARK_ROOT/tests/genai-perf-tests.json"
+  mv artifacts/ $RESULTS_FOLDER/
+
+  # upload benchmark results to buildkite
+  python3 -m pip install tabulate pandas
+  python3 "$BENCHMARK_ROOT/scripts/summary-nightly-results.py"
+  upload_to_buildkite
+
+}
+
+main "$@"

.buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh

@@ -365,7 +365,8 @@ run_serving_tests() {
       continue
     fi
 
-    server_command="$server_envs vllm serve \
+    server_command="$server_envs python3 \
+      -m vllm.entrypoints.openai.api_server \
       $server_args"
 
     # run the server
@@ -454,6 +455,11 @@ main() {
   fi
   check_hf_token
 
+  # Set to v1 to run v1 benchmark
+  if [[ "${ENGINE_VERSION:-v0}" == "v1" ]]; then
+    export VLLM_USE_V1=1
+  fi
+
   # dependencies
   (which wget && which curl) || (apt-get update && apt-get install -y wget curl)
   (which jq) || (apt-get update && apt-get -y install jq)
@@ -469,12 +475,7 @@ main() {
   ensure_sharegpt_downloaded
   declare -g RESULTS_FOLDER=results/
   mkdir -p $RESULTS_FOLDER
-  QUICK_BENCHMARK_ROOT=../.buildkite/performance-benchmarks/
-
-  # dump vllm info via vllm collect-env
-  env_output=$(vllm collect-env)
-
-  echo "$env_output" >"$RESULTS_FOLDER/vllm_env.txt"
+  QUICK_BENCHMARK_ROOT=../.buildkite/nightly-benchmarks/
 
   # benchmarking
   run_serving_tests $QUICK_BENCHMARK_ROOT/tests/"${SERVING_JSON:-serving-tests$ARCH.json}"

.buildkite/nightly-benchmarks/scripts/summary-nightly-results.py

@@ -0,0 +1,82 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import datetime
+import json
+import os
+from pathlib import Path
+
+import pandas as pd
+from tabulate import tabulate
+
+results_folder = Path("results/")
+
+# serving results and the keys that will be printed into markdown
+serving_results = []
+serving_column_mapping = {
+    "test_name": "Test name",
+    "gpu_type": "GPU",
+    "completed": "Successful req.",
+    "request_throughput": "Tput (req/s)",
+    "mean_ttft_ms": "Mean TTFT (ms)",
+    "std_ttft_ms": "Std TTFT (ms)",
+    "median_ttft_ms": "Median TTFT (ms)",
+    "mean_itl_ms": "Mean ITL (ms)",
+    "std_itl_ms": "Std ITL (ms)",
+    "median_itl_ms": "Median ITL (ms)",
+    "mean_tpot_ms": "Mean TPOT (ms)",
+    "std_tpot_ms": "Std TPOT (ms)",
+    "median_tpot_ms": "Median TPOT (ms)",
+    "total_token_throughput": "Total Token Tput (tok/s)",
+    "output_throughput": "Output Tput (tok/s)",
+    "total_input_tokens": "Total input tokens",
+    "total_output_tokens": "Total output tokens",
+    "engine": "Engine",
+}
+
+if __name__ == "__main__":
+    # collect results
+    for test_file in results_folder.glob("*.json"):
+        with open(test_file) as f:
+            raw_result = json.loads(f.read())
+
+        # attach the benchmarking command to raw_result
+        with open(test_file.with_suffix(".commands")) as f:
+            command = json.loads(f.read())
+        raw_result.update(command)
+
+        # update the test name of this result
+        raw_result.update({"test_name": test_file.stem})
+
+        # add the result to raw_result
+        serving_results.append(raw_result)
+        continue
+
+    serving_results = pd.DataFrame.from_dict(serving_results)
+
+    if not serving_results.empty:
+        serving_results = serving_results[list(serving_column_mapping.keys())].rename(
+            columns=serving_column_mapping
+        )
+
+    serving_md_table_with_headers = tabulate(
+        serving_results, headers="keys", tablefmt="pipe", showindex=False
+    )
+    # remove the first line of header
+    serving_md_table_lines = serving_md_table_with_headers.split("\n")
+    serving_md_table_without_header = "\n".join(serving_md_table_lines[2:])
+
+    prefix = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
+    prefix = prefix + "_" + os.environ.get("CURRENT_LLM_SERVING_ENGINE")
+
+    # document benchmarking results in markdown
+    with open(results_folder / f"{prefix}_nightly_results.md", "w") as f:
+        # document results with header.
+        # for those who wants to reproduce our benchmark.
+        f.write(serving_md_table_with_headers)
+        f.write("\n")
+
+    # document benchmarking results in json
+    with open(results_folder / f"{prefix}_nightly_results.json", "w") as f:
+        results = serving_results.to_dict(orient="records")
+        f.write(json.dumps(results))

.buildkite/nightly-benchmarks/scripts/wait-for-image.sh

@@ -0,0 +1,23 @@
+#!/bin/sh
+TOKEN=$(curl -s -L "https://public.ecr.aws/token?service=public.ecr.aws&scope=repository:q9t5s3a7/vllm-ci-postmerge-repo:pull" | jq -r .token)
+if [[ "$BUILDKITE_BRANCH" == "main" ]]; then
+    URL="https://public.ecr.aws/v2/q9t5s3a7/vllm-ci-postmerge-repo/manifests/$BUILDKITE_COMMIT"
+else
+    URL="https://public.ecr.aws/v2/q9t5s3a7/vllm-ci-test-repo/manifests/$BUILDKITE_COMMIT"
+fi
+
+TIMEOUT_SECONDS=10
+
+retries=0
+while [ $retries -lt 1000 ]; do
+    if [ "$(curl -s --max-time "$TIMEOUT_SECONDS" -L -H "Authorization: Bearer $TOKEN" -o /dev/null -w "%{http_code}" "$URL")" -eq 200 ]; then
+        exit 0
+    fi
+
+    echo "Waiting for image to be available..."
+
+    retries=$((retries + 1))
+    sleep 5
+done
+
+exit 1

.buildkite/nightly-benchmarks/tests/latency-tests-cpu.json

@@ -0,0 +1,30 @@
+[
+    {
+        "test_name": "latency_llama8B_tp1",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+            "load_format": "dummy",
+            "num_iters_warmup": 5,
+            "num_iters": 15
+        }
+    },
+    {
+        "test_name": "latency_llama8B_tp4",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+            "load_format": "dummy",
+            "num_iters_warmup": 5,
+            "num_iters": 15
+        }
+    }
+]

.buildkite/nightly-benchmarks/tests/serving-tests-cpu-snc3.json

@@ -95,38 +95,6 @@
             "num_prompts": 200
         }
     },
-    {
-        "test_name": "serving_llama8B_bf16_tp4_sharegpt",
-        "qps_list": ["inf"],
-        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 4,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "backend": "vllm",
-            "dataset_name": "sharegpt",
-            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
-            "num_prompts": 200
-        }
-    },
     {
         "test_name": "serving_llama8B_bf16_tp2pp3_sharegpt",
         "qps_list": ["inf"],
@@ -265,41 +233,6 @@
             "num_prompts": 1000
         }
     },
-    {
-        "test_name": "serving_llama8B_bf16_tp4_random_128_128",
-        "qps_list": ["inf"],
-        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 4,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 128,
-	    "random-output-len": 128,
-	    "ignore-eos": "",
-            "num_prompts": 1000
-        }
-    },
     {
         "test_name": "serving_llama8B_bf16_tp2pp3_random_128_128",
         "qps_list": ["inf"],
@@ -432,38 +365,6 @@
             "num_prompts": 200
         }
     },
-    {
-        "test_name": "serving_llama8B_int8_tp4_sharegpt",
-        "qps_list": ["inf"],
-        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
-            "tensor_parallel_size": 4,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
-            "backend": "vllm",
-            "dataset_name": "sharegpt",
-            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
-            "num_prompts": 200
-        }
-    },
     {
         "test_name": "serving_llama8B_int8_tp2pp3_sharegpt",
         "qps_list": ["inf"],
@@ -602,41 +503,6 @@
             "num_prompts": 1000
         }
     },
-    {
-        "test_name": "serving_llama8B_int8_tp4_random_128_128",
-        "qps_list": ["inf"],
-        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
-            "tensor_parallel_size": 4,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "RedHatAI/Meta-Llama-3.1-8B-Instruct-quantized.w8a8",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 128,
-	    "random-output-len": 128,
-	    "ignore-eos": "",
-            "num_prompts": 1000
-        }
-    },
     {
         "test_name": "serving_llama8B_int8_tp2pp3_random_128_128",
         "qps_list": ["inf"],
@@ -772,39 +638,6 @@
             "num_prompts": 200
         }
     },
-    {
-        "test_name": "serving_llama8B_int4_tp4_sharegpt",
-        "qps_list": ["inf"],
-        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
-	    "quantization": "awq",
-            "tensor_parallel_size": 4,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
-            "backend": "vllm",
-            "dataset_name": "sharegpt",
-            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
-            "num_prompts": 200
-        }
-    },
     {
         "test_name": "serving_llama8B_int4_tp2pp3_sharegpt",
         "qps_list": ["inf"],
@@ -947,42 +780,6 @@
             "num_prompts": 1000
         }
     },
-    {
-        "test_name": "serving_llama8B_int4_tp4_random_128_128",
-        "qps_list": ["inf"],
-        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200, 1000],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
-	    "quantization": "awq",
-            "tensor_parallel_size": 4,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 128,
-	    "random-output-len": 128,
-	    "ignore-eos": "",
-            "num_prompts": 1000
-        }
-    },
     {
         "test_name": "serving_llama8B_int4_tp2pp3_random_128_128",
         "qps_list": ["inf"],

.buildkite/nightly-benchmarks/tests/serving-tests-cpu.json

@@ -2,7 +2,7 @@
     {
         "test_name": "serving_llama8B_tp1_sharegpt",
         "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
         "server_environment_variables": {
             "VLLM_RPC_TIMEOUT": 100000,
 	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
@@ -28,13 +28,13 @@
             "backend": "vllm",
             "dataset_name": "sharegpt",
             "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
-            "num_prompts": 32
+            "num_prompts": 200
         }
     },
     {
         "test_name": "serving_llama8B_tp2_sharegpt",
         "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
         "server_environment_variables": {
             "VLLM_RPC_TIMEOUT": 100000,
 	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
@@ -60,13 +60,13 @@
             "backend": "vllm",
             "dataset_name": "sharegpt",
             "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
-            "num_prompts": 32
+            "num_prompts": 200
         }
     },
     {
-        "test_name": "serving_llama8B_tp1_random_128_128",
+        "test_name": "serving_llama8B_tp4_sharegpt",
         "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
         "server_environment_variables": {
             "VLLM_RPC_TIMEOUT": 100000,
 	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
@@ -76,7 +76,39 @@
         },
         "server_parameters": {
             "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 1,
+            "tensor_parallel_size": 4,
+	    "dtype": "bfloat16",
+	    "distributed_executor_backend": "mp",
+	    "block_size": 128,
+	    "trust_remote_code": "",
+            "disable_log_stats": "",
+	    "enforce_eager": "",
+	    "max_num_batched_tokens": 2048,
+	    "max_num_seqs": 256,
+            "load_format": "dummy"
+        },
+        "client_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "backend": "vllm",
+            "dataset_name": "sharegpt",
+            "dataset_path": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200
+        }
+    },
+    {
+        "test_name": "serving_llama8B_tp4_random_1024_128",
+        "qps_list": [1, 4, 16, "inf"],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
+        "server_environment_variables": {
+            "VLLM_RPC_TIMEOUT": 100000,
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
+	    "VLLM_CPU_SGL_KERNEL": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "server_parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
 	    "dtype": "bfloat16",
 	    "distributed_executor_backend": "mp",
 	    "block_size": 128,
@@ -92,16 +124,16 @@
             "model": "meta-llama/Llama-3.1-8B-Instruct",
             "backend": "vllm",
             "dataset_name": "random",
-	    "random-input-len": 128,
+	    "random-input-len": 1024,
 	    "random-output-len": 128,
 	    "ignore-eos": "",
-            "num_prompts": 32
+            "num_prompts": 100
         }
     },
     {
-        "test_name": "serving_llama8B_tp2_random_128_128",
+        "test_name": "serving_llama8B_pp6_random_1024_128",
         "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
+        "max_concurrency_list": [12, 16, 24, 32, 64, 128, 200],
         "server_environment_variables": {
             "VLLM_RPC_TIMEOUT": 100000,
 	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
@@ -111,7 +143,7 @@
         },
         "server_parameters": {
             "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 2,
+            "pipeline_parallel_size": 6,
 	    "dtype": "bfloat16",
 	    "distributed_executor_backend": "mp",
 	    "block_size": 128,
@@ -127,150 +159,10 @@
             "model": "meta-llama/Llama-3.1-8B-Instruct",
             "backend": "vllm",
             "dataset_name": "random",
-	    "random-input-len": 128,
+	    "random-input-len": 1024,
 	    "random-output-len": 128,
 	    "ignore-eos": "",
-            "num_prompts": 32
-        }
-    },
-    {
-        "test_name": "serving_llama8B_tp1_random_128_2048",
-        "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 1,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 128,
-	    "random-output-len": 2048,
-	    "ignore-eos": "",
-            "num_prompts": 32
-        }
-    },
-    {
-        "test_name": "serving_llama8B_tp2_random_128_2048",
-        "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 2,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 128,
-	    "random-output-len": 2048,
-	    "ignore-eos": "",
-            "num_prompts": 32
-        }
-    },
-    {
-        "test_name": "serving_llama8B_tp1_random_2048_128",
-        "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 1,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 2048,
-	    "random-output-len": 128,
-	    "ignore-eos": "",
-            "num_prompts": 32
-        }
-    },
-    {
-        "test_name": "serving_llama8B_tp2_random_2048_128",
-        "qps_list": [1, 4, 16, "inf"],
-        "max_concurrency_list": [32],
-        "server_environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "server_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 2,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-	    "enable_chunked_prefill": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "load_format": "dummy"
-        },
-        "client_parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "backend": "vllm",
-            "dataset_name": "random",
-	    "random-input-len": 2048,
-	    "random-output-len": 128,
-	    "ignore-eos": "",
-            "num_prompts": 32
+            "num_prompts": 100
         }
     }
 ]

.buildkite/nightly-benchmarks/tests/throughput-tests-cpu.json

@@ -0,0 +1,32 @@
+[
+    {
+        "test_name": "throughput_llama8B_tp1",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 1,
+            "load_format": "dummy",
+            "dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200,
+            "backend": "vllm"
+        }
+    },
+    {
+        "test_name": "throughput_llama8B_tp4",
+        "environment_variables": {
+	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
+	    "VLLM_CPU_KVCACHE_SPACE": 40
+        },
+        "parameters": {
+            "model": "meta-llama/Llama-3.1-8B-Instruct",
+            "tensor_parallel_size": 4,
+            "load_format": "dummy",
+            "dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
+            "num_prompts": 200,
+            "backend": "vllm"
+        }
+    }
+]

.buildkite/performance-benchmarks/tests/latency-tests-cpu.json

@@ -1,26 +0,0 @@
-[
-    {
-        "test_name": "latency_llama8B_tp2",
-        "environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 2,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "num_iters_warmup": 5,
-            "num_iters": 15
-        }
-    }
-]

.buildkite/performance-benchmarks/tests/throughput-tests-cpu.json

@@ -1,27 +0,0 @@
-[
-    {
-        "test_name": "throughput_llama8B_tp2",
-        "environment_variables": {
-            "VLLM_RPC_TIMEOUT": 100000,
-	    "VLLM_ALLOW_LONG_MAX_MODEL_LEN": 1,
-	    "VLLM_ENGINE_ITERATION_TIMEOUT_S": 120,
-	    "VLLM_CPU_SGL_KERNEL": 1,
-	    "VLLM_CPU_KVCACHE_SPACE": 40
-        },
-        "parameters": {
-            "model": "meta-llama/Llama-3.1-8B-Instruct",
-            "tensor_parallel_size": 2,
-	    "dtype": "bfloat16",
-	    "distributed_executor_backend": "mp",
-	    "block_size": 128,
-	    "trust_remote_code": "",
-            "disable_log_stats": "",
-	    "enforce_eager": "",
-	    "max_num_batched_tokens": 2048,
-	    "max_num_seqs": 256,
-            "dataset": "./ShareGPT_V3_unfiltered_cleaned_split.json",
-            "num_prompts": 200,
-            "backend": "vllm"
-        }
-    }
-]

.buildkite/pyproject.toml

@@ -0,0 +1,46 @@
+# This local pyproject file is part of the migration from yapf to ruff format.
+# It uses the same core rules as the main pyproject.toml file, but with the
+# following differences:
+# - ruff line length is overridden to 88
+# - deprecated typing ignores (UP006, UP035) have been removed
+
+[tool.ruff]
+line-length = 88
+
+[tool.ruff.lint.per-file-ignores]
+"vllm/third_party/**" = ["ALL"]
+"vllm/version.py" = ["F401"]
+"vllm/_version.py" = ["ALL"]
+
+[tool.ruff.lint]
+select = [
+    # pycodestyle
+    "E",
+    # Pyflakes
+    "F",
+    # pyupgrade
+    "UP",
+    # flake8-bugbear
+    "B",
+    # flake8-simplify
+    "SIM",
+    # isort
+    "I",
+    # flake8-logging-format
+    "G",
+]
+ignore = [
+    # star imports
+    "F405", "F403",
+    # lambda expression assignment
+    "E731",
+    # Loop control variable not used within loop body
+    "B007",
+    # f-string format
+    "UP032",
+    # Can remove once 3.10+ is the minimum Python version
+    "UP007",
+]
+
+[tool.ruff.format]
+docstring-code-format = true

.buildkite/release-pipeline.yaml

@@ -1,37 +1,24 @@
 steps:
-  # aarch64 + CUDA builds
+  # aarch64 + CUDA builds. PyTorch 2.8 aarch64 + CUDA wheel is only available on CUDA 12.9
   - label: "Build arm64 wheel - CUDA 12.9"
-    depends_on: ~
     id: build-wheel-arm64-cuda-12-9
     agents:
       queue: arm64_cpu_queue_postmerge
     commands:
       # #NOTE: torch_cuda_arch_list is derived from upstream PyTorch build files here:
       # https://github.com/pytorch/pytorch/blob/main/.ci/aarch64_linux/aarch64_ci_build.sh#L7
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg VLLM_MAIN_CUDA_VERSION=12.9 --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX 12.0' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
       - "mkdir artifacts"
       - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
       - "bash .buildkite/scripts/upload-wheels.sh"
     env:
       DOCKER_BUILDKIT: "1"
 
-  # aarch64 build
-  - label: "Build arm64 CPU wheel"
-    depends_on: ~
-    id: build-wheel-arm64-cpu
-    agents:
-      queue: arm64_cpu_queue_postmerge
-    commands:
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --build-arg VLLM_BUILD_ACL=ON --tag vllm-ci:build-image --target vllm-build --progress plain -f docker/Dockerfile.cpu ."
-      - "mkdir artifacts"
-      - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
-      - "bash .buildkite/scripts/upload-wheels.sh"
-    env:
-      DOCKER_BUILDKIT: "1"
+  - block: "Build CUDA 12.8 wheel"
+    key: block-build-cu128-wheel
 
-  # x86 + CUDA builds
   - label: "Build wheel - CUDA 12.8"
-    depends_on: ~
+    depends_on: block-build-cu128-wheel
     id: build-wheel-cuda-12-8
     agents:
       queue: cpu_queue_postmerge
@@ -43,33 +30,37 @@ steps:
     env:
       DOCKER_BUILDKIT: "1"
 
+  - block: "Build CUDA 12.6 wheel"
+    key: block-build-cu126-wheel
+    depends_on: ~
+
+  - label: "Build wheel - CUDA 12.6"
+    depends_on: block-build-cu126-wheel
+    id: build-wheel-cuda-12-6
+    agents:
+      queue: cpu_queue_postmerge
+    commands:
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.6.3 --build-arg torch_cuda_arch_list='7.0 7.5 8.0 8.9 9.0+PTX' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+      - "mkdir artifacts"
+      - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
+      - "bash .buildkite/scripts/upload-wheels.sh"
+    env:
+      DOCKER_BUILDKIT: "1"
+
+  # x86 + CUDA builds
   - label: "Build wheel - CUDA 12.9"
     depends_on: ~
     id: build-wheel-cuda-12-9
     agents:
       queue: cpu_queue_postmerge
     commands:
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='7.0 7.5 8.0 8.9 9.0+PTX' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
       - "mkdir artifacts"
       - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
       - "bash .buildkite/scripts/upload-wheels.sh"
     env:
       DOCKER_BUILDKIT: "1"
 
-  - label: "Build wheel - CUDA 13.0"
-    depends_on: ~
-    id: build-wheel-cuda-13-0
-    agents:
-      queue: cpu_queue_postmerge
-    commands:
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
-      - "mkdir artifacts"
-      - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
-      - "bash .buildkite/scripts/upload-wheels.sh"
-    env:
-      DOCKER_BUILDKIT: "1"
-
-  # Build release images (12.9)
   - label: "Build release image (x86)"
     depends_on: ~
     id: build-release-image-x86
@@ -77,12 +68,13 @@ steps:
       queue: cpu_queue_postmerge
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.8.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
       - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
       # re-tag to default image tag and push, just in case arm64 build fails
       - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
       - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
 
+  # PyTorch 2.8 aarch64 + CUDA wheel is only available on CUDA 12.9
   - label: "Build release image (arm64)"
     depends_on: ~
     id: build-release-image-arm64
@@ -90,7 +82,7 @@ steps:
       queue: arm64_cpu_queue_postmerge
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
+      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
       - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
 
   # Add job to create multi-arch manifest
@@ -110,6 +102,8 @@ steps:
     depends_on:
       - create-multi-arch-manifest
       - build-wheel-cuda-12-8
+      - build-wheel-cuda-12-6
+      - build-wheel-cuda-12-9
     id: annotate-release-workflow
     agents:
       queue: cpu_queue_postmerge
@@ -156,22 +150,6 @@ steps:
     env:
       DOCKER_BUILDKIT: "1"
 
-  - block: "Build arm64 CPU release image"
-    key: block-arm64-cpu-release-image-build
-    depends_on: ~
-
-  - label: "Build and publish arm64 CPU release image"
-    depends_on: block-arm64-cpu-release-image-build
-    agents:
-      queue: arm64_cpu_queue_postmerge
-    commands:
-      - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-      - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."
-      - "docker push public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:latest"
-      - "docker push public.ecr.aws/q9t5s3a7/vllm-arm64-cpu-release-repo:$(buildkite-agent meta-data get release-version)"
-    env:
-      DOCKER_BUILDKIT: "1"
-
   - label: "Build and publish nightly multi-arch image to DockerHub"
     depends_on:
       - create-multi-arch-manifest
@@ -180,16 +158,11 @@ steps:
       queue: cpu_queue_postmerge
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-      - "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64"
-      - "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64"
-      - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64 vllm/vllm-openai:nightly-x86_64"
-      - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64 vllm/vllm-openai:nightly-aarch64"
-      - "docker push vllm/vllm-openai:nightly-x86_64"
-      - "docker push vllm/vllm-openai:nightly-aarch64"
-      - "docker manifest create vllm/vllm-openai:nightly vllm/vllm-openai:nightly-x86_64 vllm/vllm-openai:nightly-aarch64 --amend"
-      - "docker manifest create vllm/vllm-openai:nightly-$BUILDKITE_COMMIT vllm/vllm-openai:nightly-x86_64 vllm/vllm-openai:nightly-aarch64 --amend"
-      - "docker manifest push vllm/vllm-openai:nightly"
-      - "docker manifest push vllm/vllm-openai:nightly-$BUILDKITE_COMMIT"
+      - "docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
+      - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT vllm/vllm-openai:nightly"
+      - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT vllm/vllm-openai:nightly-$BUILDKITE_COMMIT"
+      - "docker push vllm/vllm-openai:nightly"
+      - "docker push vllm/vllm-openai:nightly-$BUILDKITE_COMMIT"
       # Clean up old nightly builds (keep only last 14)
       - "bash .buildkite/scripts/cleanup-nightly-builds.sh"
     plugins:
@@ -198,4 +171,3 @@ steps:
           password-env: DOCKERHUB_TOKEN
     env:
       DOCKER_BUILDKIT: "1"
-      DOCKERHUB_USERNAME: "vllmbot"

.buildkite/scripts/annotate-release.sh

@@ -14,33 +14,18 @@ buildkite-agent annotate --style 'info' --context 'release-workflow' << EOF
 To download the wheel:
 \`\`\`
 aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux1_x86_64.whl .
-aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux2014_aarch64.whl .
-
 aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu126/vllm-${RELEASE_VERSION}+cu126-cp38-abi3-manylinux1_x86_64.whl .
-aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu129/vllm-${RELEASE_VERSION}+cu129-cp38-abi3-manylinux1_x86_64.whl .
+aws s3 cp s3://vllm-wheels/${RELEASE_VERSION}+cu118/vllm-${RELEASE_VERSION}+cu118-cp38-abi3-manylinux1_x86_64.whl . 
 \`\`\`
 
 To download and upload the image:
 
 \`\`\`
-docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64
-docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64
-
-docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64 vllm/vllm-openai:x86_64
-docker tag vllm/vllm-openai:x86_64 vllm/vllm-openai:latest-x86_64
-docker tag vllm/vllm-openai:x86_64 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64
-docker push vllm/vllm-openai:latest-x86_64
-docker push vllm/vllm-openai:v${RELEASE_VERSION}-x86_64
-
-docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64 vllm/vllm-openai:aarch64
-docker tag vllm/vllm-openai:aarch64 vllm/vllm-openai:latest-aarch64
-docker tag vllm/vllm-openai:aarch64 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
-docker push vllm/vllm-openai:latest-aarch64
-docker push vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
-
-docker manifest create vllm/vllm-openai:latest vllm/vllm-openai:latest-x86_64 vllm/vllm-openai:latest-aarch64 --amend
-docker manifest create vllm/vllm-openai:v${RELEASE_VERSION} vllm/vllm-openai:v${RELEASE_VERSION}-x86_64 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64 --amend
-docker manifest push vllm/vllm-openai:latest
-docker manifest push vllm/vllm-openai:v${RELEASE_VERSION}
+docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}
+docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT} vllm/vllm-openai
+docker tag vllm/vllm-openai vllm/vllm-openai:latest
+docker tag vllm/vllm-openai vllm/vllm-openai:v${RELEASE_VERSION}
+docker push vllm/vllm-openai:latest
+docker push vllm/vllm-openai:v${RELEASE_VERSION}
 \`\`\`
 EOF 

.buildkite/scripts/cleanup-nightly-builds.sh

@@ -8,41 +8,20 @@ set -ex
 # DockerHub API endpoint for vllm/vllm-openai repository
 REPO_API_URL="https://hub.docker.com/v2/repositories/vllm/vllm-openai/tags"
 
-# Get DockerHub credentials from environment
+# Get DockerHub token from environment
 if [ -z "$DOCKERHUB_TOKEN" ]; then
     echo "Error: DOCKERHUB_TOKEN environment variable is not set"
     exit 1
 fi
 
-if [ -z "$DOCKERHUB_USERNAME" ]; then
-    echo "Error: DOCKERHUB_USERNAME environment variable is not set"
-    exit 1
-fi
-
-# Get DockerHub bearer token
-echo "Getting DockerHub bearer token..."
-set +x
-BEARER_TOKEN=$(curl -s -X POST \
-    -H "Content-Type: application/json" \
-    -d "{\"username\": \"$DOCKERHUB_USERNAME\", \"password\": \"$DOCKERHUB_TOKEN\"}" \
-    "https://hub.docker.com/v2/users/login" | jq -r '.token')
-set -x
-
-if [ -z "$BEARER_TOKEN" ] || [ "$BEARER_TOKEN" = "null" ]; then
-    echo "Error: Failed to get DockerHub bearer token"
-    exit 1
-fi
-
 # Function to get all tags from DockerHub
 get_all_tags() {
     local page=1
     local all_tags=""
     
     while true; do
-        set +x
-        local response=$(curl -s -H "Authorization: Bearer $BEARER_TOKEN" \
+        local response=$(curl -s -H "Authorization: Bearer $DOCKERHUB_TOKEN" \
             "$REPO_API_URL?page=$page&page_size=100")
-        set -x
         
         # Get both last_updated timestamp and tag name, separated by |
         local tags=$(echo "$response" | jq -r '.results[] | select(.name | startswith("nightly-")) | "\(.last_updated)|\(.name)"')
@@ -64,9 +43,7 @@ delete_tag() {
     echo "Deleting tag: $tag_name"
     
     local delete_url="https://hub.docker.com/v2/repositories/vllm/vllm-openai/tags/$tag_name"
-    set +x
-    local response=$(curl -s -X DELETE -H "Authorization: Bearer $BEARER_TOKEN" "$delete_url")
-    set -x
+    local response=$(curl -s -X DELETE -H "Authorization: Bearer $DOCKERHUB_TOKEN" "$delete_url")
     
     if echo "$response" | jq -e '.detail' > /dev/null 2>&1; then
         echo "Warning: Failed to delete tag $tag_name: $(echo "$response" | jq -r '.detail')"

.buildkite/scripts/hardware_ci/run-amd-test.sh

@@ -86,6 +86,10 @@ if [[ $commands == *"pytest -v -s models/test_registry.py"* ]]; then
   commands=${commands//"pytest -v -s models/test_registry.py"/"pytest -v -s models/test_registry.py -k 'not BambaForCausalLM and not GritLM and not Mamba2ForCausalLM and not Zamba2ForCausalLM'"}
 fi
 
+if [[ $commands == *"VLLM_USE_V1=0 pytest -v -s models/test_initialization.py -k 'not llama4 and not plamo2'"* ]]; then
+  commands=${commands//"VLLM_USE_V1=0 pytest -v -s models/test_initialization.py -k 'not llama4 and not plamo2'"/"VLLM_USE_V1=0 pytest -v -s models/test_initialization.py -k 'not llama4 and not plamo2 and not BambaForCausalLM and not Gemma2ForCausalLM and not Grok1ModelForCausalLM and not Zamba2ForCausalLM and not Gemma2Model and not GritLM'"}
+fi
+
 if [[ $commands == *"pytest -v -s compile/test_basic_correctness.py"* ]]; then
   commands=${commands//"pytest -v -s compile/test_basic_correctness.py"/"VLLM_USE_TRITON_FLASH_ATTN=0 pytest -v -s compile/test_basic_correctness.py"}
 fi
@@ -163,6 +167,12 @@ if [[ $commands == *" entrypoints/llm "* ]]; then
   --ignore=entrypoints/llm/test_prompt_validation.py "}
 fi
 
+#Obsolete currently
+##ignore certain Entrypoints/llm tests
+#if [[ $commands == *" && pytest -v -s entrypoints/llm/test_guided_generate.py"* ]]; then
+#  commands=${commands//" && pytest -v -s entrypoints/llm/test_guided_generate.py"/" "}
+#fi
+
 # --ignore=entrypoints/openai/test_encoder_decoder.py \
 # --ignore=entrypoints/openai/test_embedding.py \
 # --ignore=entrypoints/openai/test_oot_registration.py

.buildkite/scripts/hardware_ci/run-cpu-test-ppc64le.sh

@@ -25,28 +25,25 @@ function cpu_tests() {
 
   # offline inference
   podman exec -it "$container_id" bash -c "
-    set -xve
-    python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m" >> $HOME/test_basic.log
+    set -e
+    python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m"
 
   # Run basic model test
   podman exec -it "$container_id" bash -c "
-    set -evx
+    set -e
     pip install pytest pytest-asyncio einops peft Pillow soundfile transformers_stream_generator matplotlib
     pip install sentence-transformers datamodel_code_generator
-
-    # Note: disable Bart until supports V1
-    # pytest -v -s tests/models/language/generation/test_bart.py -m cpu_model
+    pytest -v -s tests/models/language/generation/test_bart.py -m cpu_model
     pytest -v -s tests/models/language/generation/test_common.py::test_models[False-5-32-openai-community/gpt2]
     pytest -v -s tests/models/language/generation/test_common.py::test_models[False-5-32-facebook/opt-125m]
     pytest -v -s tests/models/language/generation/test_common.py::test_models[False-5-32-google/gemma-1.1-2b-it]
     pytest -v -s tests/models/language/pooling/test_classification.py::test_models[float-jason9693/Qwen2.5-1.5B-apeach]
-    # TODO: Below test case tests/models/language/pooling/test_embedding.py::test_models[True-ssmits/Qwen2-7B-Instruct-embed-base] fails on ppc64le. Disabling it for time being.
-    # pytest -v -s tests/models/language/pooling/test_embedding.py -m cpu_model" >> $HOME/test_rest.log
+    pytest -v -s tests/models/language/pooling/test_embedding.py -m cpu_model"
 }
 
 # All of CPU tests are expected to be finished less than 40 mins.
 
 export container_id
 export -f cpu_tests
-timeout 120m bash -c cpu_tests
+timeout 40m bash -c cpu_tests
 

.buildkite/scripts/hardware_ci/run-cpu-test.sh

@@ -58,11 +58,15 @@ function cpu_tests() {
     # pytest -x -v -s tests/kernels/attention/test_cache.py -m cpu_model
     # pytest -x -v -s tests/kernels/attention/test_mla_decode_cpu.py -m cpu_model
 
-    pytest -x -v -s tests/models/language/generation -m cpu_model
-    VLLM_CPU_SGL_KERNEL=1 pytest -x -v -s tests/models/language/generation -m cpu_model
+    # Note: disable Bart until supports V1
+    pytest -x -v -s tests/models/language/generation -m cpu_model \
+                --ignore=tests/models/language/generation/test_bart.py
+    VLLM_CPU_SGL_KERNEL=1 pytest -x -v -s tests/models/language/generation -m cpu_model \
+                --ignore=tests/models/language/generation/test_bart.py
 
     pytest -x -v -s tests/models/language/pooling -m cpu_model
     pytest -x -v -s tests/models/multimodal/generation \
+                --ignore=tests/models/multimodal/generation/test_mllama.py \
                 --ignore=tests/models/multimodal/generation/test_pixtral.py \
                 -m cpu_model"
 
@@ -70,7 +74,7 @@ function cpu_tests() {
   docker exec cpu-test-"$NUMA_NODE" bash -c "
     set -e
     pytest -x -s -v \
-    tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_logprobs"
+    tests/quantization/test_compressed_tensors.py::test_compressed_tensors_w8a8_logprobs[False-10-32-neuralmagic/Llama-3.2-1B-quantized.w8a8]"
 
   # Note: disable it until supports V1
   # Run AWQ test

.buildkite/scripts/hardware_ci/run-npu-test.sh

@@ -1,191 +0,0 @@
-#!/bin/bash
-
-# This script build the Ascend NPU docker image and run the offline inference inside the container.
-# It serves a sanity check for compilation and basic model usage.
-set -ex
-
-# Base ubuntu image with basic ascend development libraries and python installed
-VLLM_ASCEND_REPO="https://github.com/vllm-project/vllm-ascend.git"
-CONFIG_FILE_REMOTE_PATH="tests/e2e/vllm_interface/vllm_test.cfg"
-TEST_RUN_CONFIG_FILE="vllm_test.cfg"
-VLLM_ASCEND_TMP_DIR=
-# Get the test run configuration file from the vllm-ascend repository
-fetch_vllm_test_cfg() {
-    VLLM_ASCEND_TMP_DIR=$(mktemp -d)
-    # Ensure that the temporary directory is cleaned up when an exception occurs during configuration file retrieval
-    cleanup() {
-        rm -rf "${VLLM_ASCEND_TMP_DIR}"
-    }
-    trap cleanup EXIT
-
-    GIT_TRACE=1 git clone -v --depth 1 "${VLLM_ASCEND_REPO}" "${VLLM_ASCEND_TMP_DIR}"
-    if [ ! -f "${VLLM_ASCEND_TMP_DIR}/${CONFIG_FILE_REMOTE_PATH}" ]; then
-        echo "Error: file '${CONFIG_FILE_REMOTE_PATH}' does not exist in the warehouse" >&2
-        exit 1
-    fi
-
-    # If the file already exists locally, just overwrite it
-    cp "${VLLM_ASCEND_TMP_DIR}/${CONFIG_FILE_REMOTE_PATH}" "${TEST_RUN_CONFIG_FILE}"
-    echo "Copied ${CONFIG_FILE_REMOTE_PATH} to ${TEST_RUN_CONFIG_FILE}"
-
-    # Since the trap will be overwritten later, and when it is executed here, the task of cleaning up resources
-    # when the trap is abnormal has been completed, so the temporary resources are manually deleted here.
-    rm -rf "${VLLM_ASCEND_TMP_DIR}"
-    trap - EXIT
-}
-
-# Downloads test run configuration file from a remote URL.
-# Loads the configuration into the current script environment.
-get_config() {
-    if [ ! -f "${TEST_RUN_CONFIG_FILE}" ]; then
-        echo "Error: file '${TEST_RUN_CONFIG_FILE}' does not exist in the warehouse" >&2
-        exit 1
-    fi
-    source "${TEST_RUN_CONFIG_FILE}"
-    echo "Base docker image name that get from configuration: ${BASE_IMAGE_NAME}"
-    return 0
-}
-
-# get test running configuration.
-fetch_vllm_test_cfg
-get_config
-# Check if the function call was successful. If not, exit the script.
-if [ $? -ne 0 ]; then
-  exit 1
-fi
-
-image_name="npu/vllm-ci:${BUILDKITE_COMMIT}_${EPOCHSECONDS}"
-container_name="npu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head -c 10; echo)"
-
-# BUILDKITE_AGENT_NAME format is {hostname}-{agent_idx}-{npu_card_num}cards
-agent_idx=$(echo "${BUILDKITE_AGENT_NAME}" | awk -F'-' '{print $(NF-1)}')
-echo "agent_idx: ${agent_idx}"
-builder_name="cachebuilder${agent_idx}"
-builder_cache_dir="/mnt/docker-cache${agent_idx}"
-mkdir -p ${builder_cache_dir}
-
-# Try building the docker image
-cat <<EOF | DOCKER_BUILDKIT=1 docker build \
-    --add-host cache-service-vllm.nginx-pypi-cache.svc.cluster.local:${PYPI_CACHE_HOST} \
-    --builder ${builder_name} --cache-from type=local,src=${builder_cache_dir} \
-                           --cache-to type=local,dest=${builder_cache_dir},mode=max \
-    --progress=plain --load -t ${image_name} -f - .
-FROM ${BASE_IMAGE_NAME}
-
-# Define environments
-ENV DEBIAN_FRONTEND=noninteractive
-
-RUN pip config set global.index-url http://cache-service-vllm.nginx-pypi-cache.svc.cluster.local:${PYPI_CACHE_PORT}/pypi/simple && \
-    pip config set global.trusted-host cache-service-vllm.nginx-pypi-cache.svc.cluster.local && \
-    apt-get update -y && \
-    apt-get install -y python3-pip git vim wget net-tools gcc g++ cmake libnuma-dev && \
-    rm -rf /var/cache/apt/* && \
-    rm -rf /var/lib/apt/lists/*
-
-# Install for pytest to make the docker build cache layer always valid
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install pytest>=6.0  modelscope
-
-WORKDIR /workspace/vllm
-
-# Install vLLM dependencies in advance. Effect: As long as common.txt remains unchanged, the docker cache layer will be valid.
-COPY requirements/common.txt /workspace/vllm/requirements/common.txt
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install -r requirements/common.txt
-
-COPY . .
-
-# Install vLLM
-RUN --mount=type=cache,target=/root/.cache/pip \
-    VLLM_TARGET_DEVICE="empty" python3 -m pip install -v -e /workspace/vllm/ --extra-index https://download.pytorch.org/whl/cpu/ && \
-    python3 -m pip uninstall -y triton
-
-# Install vllm-ascend
-WORKDIR /workspace
-ARG VLLM_ASCEND_REPO=https://github.com/vllm-project/vllm-ascend.git
-ARG VLLM_ASCEND_TAG=main
-RUN git config --global url."https://gh-proxy.test.osinfra.cn/https://github.com/".insteadOf "https://github.com/" && \
-    git clone --depth 1 \$VLLM_ASCEND_REPO --branch \$VLLM_ASCEND_TAG /workspace/vllm-ascend
-
-# Install vllm dependencies in advance. Effect: As long as common.txt remains unchanged, the docker cache layer will be valid.
-RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install -r /workspace/vllm-ascend/requirements.txt
-
-RUN --mount=type=cache,target=/root/.cache/pip \
-    export PIP_EXTRA_INDEX_URL=https://mirrors.huaweicloud.com/ascend/repos/pypi && \
-    source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
-    source /usr/local/Ascend/nnal/atb/set_env.sh && \
-    export LD_LIBRARY_PATH=\$LD_LIBRARY_PATH:/usr/local/Ascend/ascend-toolkit/latest/`uname -i`-linux/devlib && \
-    python3 -m pip install -v -e /workspace/vllm-ascend/ --extra-index https://download.pytorch.org/whl/cpu/
-
-ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
-ENV VLLM_USE_MODELSCOPE=True
-
-WORKDIR /workspace/vllm-ascend
-
-CMD ["/bin/bash"]
-
-EOF
-
-# Setup cleanup
-remove_docker_container() {
-  docker rm -f "${container_name}" || true;
-  docker image rm -f "${image_name}" || true;
-  docker system prune -f || true;
-}
-trap remove_docker_container EXIT
-
-# Generate corresponding --device args based on BUILDKITE_AGENT_NAME
-# Ascend NPU BUILDKITE_AGENT_NAME format is {hostname}-{agent_idx}-{npu_card_num}cards, and agent_idx starts from 1.
-#   e.g. atlas-a2-001-1-2cards means this is the 1-th agent on atlas-a2-001 host, and it has 2 NPU cards.
-#   returns --device /dev/davinci0 --device /dev/davinci1
-parse_and_gen_devices() {
-    local input="$1"
-    local index cards_num
-    if [[ "$input" =~ ([0-9]+)-([0-9]+)cards$ ]]; then
-        index="${BASH_REMATCH[1]}"
-        cards_num="${BASH_REMATCH[2]}"
-    else
-        echo "parse error" >&2
-        return 1
-    fi
-
-    local devices=""
-    local i=0
-    while (( i < cards_num )); do
-        local dev_idx=$(((index - 1)*cards_num + i ))
-        devices="$devices --device /dev/davinci${dev_idx}"
-        ((i++))
-    done
-
-    # trim leading space
-    devices="${devices#"${devices%%[![:space:]]*}"}"
-    # Output devices: assigned to the caller variable
-    printf '%s' "$devices"
-}
-
-devices=$(parse_and_gen_devices "${BUILDKITE_AGENT_NAME}") || exit 1
-
-# Run the image and execute the Out-Of-Tree (OOT) platform interface test case on Ascend NPU hardware.
-# This test checks whether the OOT platform interface is functioning properly in conjunction with
-# the hardware plugin vllm-ascend.
-model_cache_dir=/mnt/modelscope${agent_idx}
-mkdir -p ${model_cache_dir}
-docker run \
-    ${devices} \
-    --device /dev/davinci_manager \
-    --device /dev/devmm_svm \
-    --device /dev/hisi_hdc \
-    -v /usr/local/dcmi:/usr/local/dcmi \
-    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
-    -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
-    -v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
-    -v /etc/ascend_install.info:/etc/ascend_install.info \
-    -v ${model_cache_dir}:/root/.cache/modelscope \
-    --entrypoint="" \
-    --name "${container_name}" \
-    "${image_name}" \
-    bash -c '
-    set -e
-    pytest -v -s tests/e2e/vllm_interface/
-'

.buildkite/scripts/hardware_ci/run-tpu-v1-test-part2.sh

@@ -62,11 +62,12 @@ echo "--- Installing Python dependencies ---"
 python3 -m pip install --progress-bar off git+https://github.com/thuml/depyf.git \
     && python3 -m pip install --progress-bar off pytest pytest-asyncio tpu-info \
     && python3 -m pip install --progress-bar off "lm-eval @ git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d" \
-    && python3 -m pip install --progress-bar off hf-transfer tblib==3.1.0
+    && python3 -m pip install --progress-bar off hf-transfer
 echo "--- Python dependencies installed ---"
-
+export VLLM_USE_V1=1
 export VLLM_XLA_CHECK_RECOMPILATION=1
 export VLLM_XLA_CACHE_PATH=
+echo "Using VLLM V1"
 
 echo "--- Hardware Information ---"
 # tpu-info

.buildkite/scripts/hardware_ci/run-tpu-v1-test.sh

@@ -62,11 +62,12 @@ echo "--- Installing Python dependencies ---"
 python3 -m pip install --progress-bar off git+https://github.com/thuml/depyf.git \
     && python3 -m pip install --progress-bar off pytest pytest-asyncio tpu-info \
     && python3 -m pip install --progress-bar off "lm-eval @ git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d" \
-    && python3 -m pip install --progress-bar off hf-transfer tblib==3.1.0
+    && python3 -m pip install --progress-bar off hf-transfer
 echo "--- Python dependencies installed ---"
-
+export VLLM_USE_V1=1
 export VLLM_XLA_CHECK_RECOMPILATION=1
 export VLLM_XLA_CACHE_PATH=
+echo "Using VLLM V1"
 
 echo "--- Hardware Information ---"
 # tpu-info

.buildkite/scripts/hardware_ci/run-xpu-test.sh

@@ -20,10 +20,7 @@ trap remove_docker_container EXIT
 
 # Run the image and test offline inference/tensor parallel
 docker run \
-    --device /dev/dri:/dev/dri \
-    --net=host \
-    --ipc=host \
-    --privileged \
+    --device /dev/dri \
     -v /dev/dri/by-path:/dev/dri/by-path \
     --entrypoint="" \
     -e "HF_TOKEN=${HF_TOKEN}" \
@@ -33,19 +30,20 @@ docker run \
     bash -c '
     set -e
     echo $ZE_AFFINITY_MASK
-    pip install tblib==3.1.0
     python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
     python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 -O3 -O.cudagraph_mode=NONE
     python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend ray
     python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager -tp 2 --distributed-executor-backend mp
-    VLLM_ATTENTION_BACKEND=TRITON_ATTN python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
+    VLLM_ATTENTION_BACKEND=TRITON_ATTN_VLLM_V1 python3 examples/offline_inference/basic/generate.py --model facebook/opt-125m --block-size 64 --enforce-eager
     cd tests
     pytest -v -s v1/core
     pytest -v -s v1/engine
     pytest -v -s v1/sample --ignore=v1/sample/test_logprobs.py --ignore=v1/sample/test_logprobs_e2e.py
     pytest -v -s v1/worker --ignore=v1/worker/test_gpu_model_runner.py
     pytest -v -s v1/structured_output
-    pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_tree_attention.py --ignore=v1/spec_decode/test_speculators_eagle3.py
+    pytest -v -s v1/spec_decode --ignore=v1/spec_decode/test_max_len.py --ignore=v1/spec_decode/test_eagle.py --ignore=v1/spec_decode/test_tree_attention.py
     pytest -v -s v1/kv_connector/unit --ignore=v1/kv_connector/unit/test_multi_connector.py --ignore=v1/kv_connector/unit/test_nixl_connector.py --ignore=v1/kv_connector/unit/test_shared_storage_connector.py
     pytest -v -s v1/test_serial_utils.py
+    pytest -v -s v1/test_utils.py
+    pytest -v -s v1/test_metrics_reader.py
 '

.buildkite/scripts/run-benchmarks.sh

@@ -18,7 +18,7 @@ vllm bench throughput --input-len 256 --output-len 256 --output-json throughput_
 bench_throughput_exit_code=$?
 
 # run server-based benchmarks and upload the result to buildkite
-vllm serve meta-llama/Llama-2-7b-chat-hf &
+python3 -m vllm.entrypoints.openai.api_server --model meta-llama/Llama-2-7b-chat-hf &
 server_pid=$!
 wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
 

.buildkite/scripts/run-prime-rl-test.sh

@@ -1,59 +0,0 @@
-#!/bin/bash
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-
-# Setup script for Prime-RL integration tests
-# This script prepares the environment for running Prime-RL tests with nightly vLLM
-
-set -euo pipefail
-
-SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-REPO_ROOT="$(cd "${SCRIPT_DIR}/../.." && pwd)"
-PRIME_RL_REPO="https://github.com/PrimeIntellect-ai/prime-rl.git"
-PRIME_RL_DIR="${REPO_ROOT}/prime-rl"
-
-echo "Setting up Prime-RL integration test environment..."
-
-# Clean up any existing Prime-RL directory
-if [ -d "${PRIME_RL_DIR}" ]; then
-    echo "Removing existing Prime-RL directory..."
-    rm -rf "${PRIME_RL_DIR}"
-fi
-
-# Install UV if not available
-if ! command -v uv &> /dev/null; then
-    echo "Installing UV package manager..."
-    curl -LsSf https://astral.sh/uv/install.sh | sh
-    source $HOME/.local/bin/env
-fi
-
-# Clone Prime-RL repository at specific branch for reproducible tests
-PRIME_RL_BRANCH="integ-vllm-main"
-echo "Cloning Prime-RL repository at branch: ${PRIME_RL_BRANCH}..."
-git clone --branch "${PRIME_RL_BRANCH}" --single-branch "${PRIME_RL_REPO}" "${PRIME_RL_DIR}"
-cd "${PRIME_RL_DIR}"
-
-echo "Setting up UV project environment..."
-export UV_PROJECT_ENVIRONMENT=/usr/local
-ln -s /usr/bin/python3 /usr/local/bin/python
-
-# Remove vllm pin from pyproject.toml
-echo "Removing vllm pin from pyproject.toml..."
-sed -i '/vllm==/d' pyproject.toml
-
-# Sync Prime-RL dependencies
-echo "Installing Prime-RL dependencies..."
-uv sync --inexact && uv sync --inexact --all-extras
-
-# Verify installation
-echo "Verifying installations..."
-uv run python -c "import vllm; print(f'vLLM version: {vllm.__version__}')"
-uv run python -c "import prime_rl; print('Prime-RL imported successfully')"
-
-echo "Prime-RL integration test environment setup complete!"
-
-echo "Running Prime-RL integration tests..."
-export WANDB_MODE=offline # this makes this test not require a WANDB_API_KEY
-uv run pytest -vs tests/integration/test_rl.py -m gpu
-
-echo "Prime-RL integration tests completed!"

.buildkite/scripts/tpu/quantized_v6e_1.env

@@ -9,6 +9,6 @@ MAX_NUM_BATCHED_TOKENS=1024
 TENSOR_PARALLEL_SIZE=1
 MAX_MODEL_LEN=2048
 DOWNLOAD_DIR=/mnt/disks/persist
-EXPECTED_THROUGHPUT=8.7
+EXPECTED_THROUGHPUT=10.0
 INPUT_LEN=1800
 OUTPUT_LEN=128

.buildkite/scripts/tpu/run_bm.sh

@@ -42,7 +42,7 @@ echo "lanching vllm..."
 echo "logging to $VLLM_LOG"
 echo
 
-vllm serve $MODEL \
+VLLM_USE_V1=1 vllm serve $MODEL \
  --seed 42 \
  --max-num-seqs $MAX_NUM_SEQS \
  --max-num-batched-tokens $MAX_NUM_BATCHED_TOKENS \

.buildkite/scripts/upload-wheels.sh

@@ -58,25 +58,33 @@ python3 .buildkite/generate_index.py --wheel "$normal_wheel"
 aws s3 cp "$wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
 aws s3 cp "$normal_wheel" "s3://vllm-wheels/$BUILDKITE_COMMIT/"
 
-if [[ $normal_wheel == *"cu129"* ]]; then
+if [[ $normal_wheel == *"cu126"* ]]; then
+    # if $normal_wheel matches cu126, do not upload the index.html
+    echo "Skipping index files for cu126 wheels"
+elif [[ $normal_wheel == *"cu128"* ]]; then
+    # if $normal_wheel matches cu128, do not upload the index.html
+    echo "Skipping index files for cu128 wheels"
+else
     # only upload index.html for cu129 wheels (default wheels) as it
     # is available on both x86 and arm64
     aws s3 cp index.html "s3://vllm-wheels/$BUILDKITE_COMMIT/vllm/index.html"
     aws s3 cp "s3://vllm-wheels/nightly/index.html" "s3://vllm-wheels/$BUILDKITE_COMMIT/index.html"
-else
-    echo "Skipping index files for non-cu129 wheels"
 fi
 
 # generate index for nightly
 aws s3 cp "$wheel" "s3://vllm-wheels/nightly/"
 aws s3 cp "$normal_wheel" "s3://vllm-wheels/nightly/"
 
-if [[ $normal_wheel == *"cu129"* ]]; then
+if [[ $normal_wheel == *"cu126"* ]]; then
+    # if $normal_wheel matches cu126, do not upload the index.html
+    echo "Skipping index files for cu126 wheels"
+elif [[ $normal_wheel == *"cu128"* ]]; then
+    # if $normal_wheel matches cu128, do not upload the index.html
+    echo "Skipping index files for cu128 wheels"
+else
     # only upload index.html for cu129 wheels (default wheels) as it
     # is available on both x86 and arm64
     aws s3 cp index.html "s3://vllm-wheels/nightly/vllm/index.html"
-else
-    echo "Skipping index files for non-cu129 wheels"
 fi
 
 aws s3 cp "$wheel" "s3://vllm-wheels/$version/"

.buildkite/test-pipeline.yaml

@@ -6,28 +6,24 @@
 # to generate the final pipeline yaml file.
 
 # Documentation
-# label(str): the name of the test. emojis allowed.
-# fast_check(bool): whether to run this on each commit on the fastcheck pipeline.
-# torch_nightly(bool): whether to run this on vllm against the torch nightly pipeline.
-# fast_check_only(bool): run this test on the fastcheck pipeline only
-# optional(bool): never run this test by default (i.e. need to unblock manually) unless it's a scheduled nightly run.
-# soft_fail(bool): allow this step to fail without failing the entire pipeline (useful for flaky or experimental tests).
+# label(str): the name of the test. emoji allowed.
+# fast_check(bool): whether to run this on each commit on fastcheck pipeline.
+# torch_nightly(bool): whether to run this on vllm against torch nightly pipeline.
+# fast_check_only(bool): run this test on fastcheck pipeline only
+# optional(bool): never run this test by default (i.e. need to unblock manually) unless it's scheduled nightly run.
 # command(str): the single command to run for tests. incompatible with commands.
-# commands(list): the list of commands to run for the test. incompatible with command.
-# mirror_hardwares(list): the list of hardware to run the test on as well. currently only supports [amdexperimental]
-# gpu(str): override the GPU selection for the test. default is L4 GPUs. supports a100, b200, h200
-# num_gpus(int): override the number of GPUs for the test. defaults to 1 GPU. currently supports 2,4.
-# num_nodes(int): whether to simulate multi-node setup by launching multiple containers on one host,
-#     in this case, commands must be specified. the first command runs on the first host, the second
+# commands(list): the list of commands to run for test. incompatbile with command.
+# mirror_hardwares(list): the list of hardwares to run the test on as well. currently only supports [amd]
+# gpu(str): override the GPU selection for the test. default is on L4 GPUs. currently only supports a100
+# num_gpus(int): override the number of GPUs for the test. default to 1 GPU. currently support 2,4.
+# num_nodes(int): whether to simulate multi-node setup by launch multiple containers on one host,
+#     in this case, commands must be specified. the first command runs on first host, the second
 #     command runs on the second host.
-# timeout_in_minutes(int): sets a timeout for the step in minutes. if not specified, uses the default timeout.
-# parallelism(int): number of parallel jobs to run for this step. enables test sharding using $$BUILDKITE_PARALLEL_JOB
-#     and $$BUILDKITE_PARALLEL_JOB_COUNT environment variables.
-# working_dir(str): specify the place where the command should execute, default to /vllm-workspace/tests
-# source_file_dependencies(list): the list of prefixes to opt-in the test for, if empty, the test will always run.
+# working_dir(str): specify the place where command should execute, default to /vllm-workspace/tests
+# source_file_dependencies(list): the list of prefix to opt-in the test for, if empty, the test will always run.
 
 # When adding a test
-# - If the test belongs to an existing group, add it there
+# - If the test belong to an existing group, add it there
 # - If the test is short, add to any existing step
 # - If the test takes more than 10min, then it is okay to create a new step.
 #   Note that all steps execute in parallel.
@@ -38,7 +34,7 @@ steps:
 - label: Pytorch Nightly Dependency Override Check # 2min
   # if this test fails, it means the nightly torch version is not compatible with some
   # of the dependencies. Please check the error message and add the package to whitelist
-  # in /vllm/tools/pre_commit/generate_nightly_torch_test.py
+  # in /vllm/tools/generate_nightly_torch_test.py
   soft_fail: true
   source_file_dependencies:
   - requirements/nightly_torch_test.txt
@@ -50,28 +46,23 @@ steps:
   mirror_hardwares: [amdexperimental]
   source_file_dependencies:
   - vllm/
-  - tests/multimodal
-  - tests/utils_
-  commands:
-  - pytest -v -s -m 'not cpu_test' multimodal
-  - pytest -v -s utils_
-
-- label: Async Engine, Inputs, Utils, Worker Test (CPU) # 4 mins
-  timeout_in_minutes: 10
-  source_file_dependencies:
-  - vllm/
+  - tests/mq_llm_engine
+  - tests/async_engine
   - tests/test_inputs.py
   - tests/test_outputs.py
   - tests/multimodal
+  - tests/utils_
+  - tests/worker
   - tests/standalone_tests/lazy_imports.py
-  - tests/transformers_utils
-  no_gpu: true
   commands:
   - python3 standalone_tests/lazy_imports.py
+  - pytest -v -s mq_llm_engine # MQLLMEngine
+  - pytest -v -s async_engine # AsyncLLMEngine
   - pytest -v -s test_inputs.py
   - pytest -v -s test_outputs.py
-  - pytest -v -s -m 'cpu_test' multimodal
-  - pytest -v -s transformers_utils
+  - pytest -v -s multimodal
+  - pytest -v -s utils_ # Utils
+  - pytest -v -s worker # Worker
 
 - label: Python-only Installation Test # 10min
   timeout_in_minutes: 20
@@ -91,25 +82,27 @@ steps:
   - vllm/
   - tests/basic_correctness/test_basic_correctness
   - tests/basic_correctness/test_cpu_offload
+  - tests/basic_correctness/test_preemption
   - tests/basic_correctness/test_cumem.py
   commands:
   - export VLLM_WORKER_MULTIPROC_METHOD=spawn
   - pytest -v -s basic_correctness/test_cumem.py
   - pytest -v -s basic_correctness/test_basic_correctness.py
   - pytest -v -s basic_correctness/test_cpu_offload.py
+  - VLLM_TEST_ENABLE_ARTIFICIAL_PREEMPT=1 pytest -v -s basic_correctness/test_preemption.py
 
-- label: Entrypoints Unit Tests # 5min
-  timeout_in_minutes: 10
-  working_dir: "/vllm-workspace/tests"
+- label: Core Test # 22min
+  timeout_in_minutes: 35
+  mirror_hardwares: [amdexperimental]
   fast_check: true
   source_file_dependencies:
-  - vllm/entrypoints
-  - tests/entrypoints/
+  - vllm/core
+  - vllm/distributed
+  - tests/core
   commands:
-  - pytest -v -s entrypoints/openai/tool_parsers
-  - pytest -v -s entrypoints/ --ignore=entrypoints/llm --ignore=entrypoints/openai --ignore=entrypoints/offline_mode --ignore=entrypoints/test_chat_utils.py  --ignore=entrypoints/pooling
+  - pytest -v -s core
 
-- label: Entrypoints Integration Test (LLM) # 30min
+- label: Entrypoints Test (LLM) # 30min
   timeout_in_minutes: 40
   mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
@@ -121,11 +114,12 @@ steps:
   - tests/entrypoints/offline_mode
   commands:
   - export VLLM_WORKER_MULTIPROC_METHOD=spawn
-  - pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_collective_rpc.py
+  - pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_lazy_outlines.py --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_collective_rpc.py
+  - pytest -v -s entrypoints/llm/test_lazy_outlines.py # it needs a clean process
   - pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
-  - pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
+  - VLLM_USE_V1=0 pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
 
-- label: Entrypoints Integration Test (API Server) # 100min
+- label: Entrypoints Test (API Server) # 100min
   timeout_in_minutes: 130
   mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
@@ -138,22 +132,9 @@ steps:
   commands:
   - export VLLM_WORKER_MULTIPROC_METHOD=spawn
   - PYTHONPATH=/vllm-workspace pytest -v -s entrypoints/openai/test_collective_rpc.py # PYTHONPATH is needed to import custom Worker extension
-  - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_chat_with_tool_reasoning.py --ignore=entrypoints/openai/test_oot_registration.py --ignore=entrypoints/openai/test_tensorizer_entrypoint.py --ignore=entrypoints/openai/correctness/ --ignore=entrypoints/openai/test_collective_rpc.py --ignore=entrypoints/openai/tool_parsers/
+  - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_chat_with_tool_reasoning.py --ignore=entrypoints/openai/test_oot_registration.py --ignore=entrypoints/openai/test_tensorizer_entrypoint.py --ignore=entrypoints/openai/correctness/ --ignore=entrypoints/openai/test_collective_rpc.py
   - pytest -v -s entrypoints/test_chat_utils.py
 
-- label: Entrypoints Integration Test (Pooling)
-  timeout_in_minutes: 50
-  mirror_hardwares: [amdexperimental]
-  working_dir: "/vllm-workspace/tests"
-  fast_check: true
-  torch_nightly: true
-  source_file_dependencies:
-  - vllm/
-  - tests/entrypoints/pooling
-  commands:
-  - export VLLM_WORKER_MULTIPROC_METHOD=spawn
-  - pytest -v -s entrypoints/pooling
-
 - label: Distributed Tests (4 GPUs) # 35min
   timeout_in_minutes: 50
   mirror_hardwares: [amdexperimental]
@@ -161,6 +142,7 @@ steps:
   num_gpus: 4
   source_file_dependencies:
   - vllm/distributed/
+  - vllm/core/
   - tests/distributed/test_utils
   - tests/distributed/test_pynccl
   - tests/distributed/test_events
@@ -168,36 +150,28 @@ steps:
   - examples/offline_inference/rlhf.py
   - examples/offline_inference/rlhf_colocate.py
   - tests/examples/offline_inference/data_parallel.py
-  - tests/v1/distributed
+  - tests/v1/test_async_llm_dp.py
+  - tests/v1/test_external_lb_dp.py
+  - tests/v1/test_internal_lb_dp.py
+  - tests/v1/test_hybrid_lb_dp.py
   - tests/v1/engine/test_engine_core_client.py
-  - tests/distributed/test_symm_mem_allreduce.py
   commands:
-  # https://github.com/NVIDIA/nccl/issues/1838
-  - export NCCL_CUMEM_HOST_ENABLE=0
-  # test with torchrun tp=2 and external_dp=2
+  # test with tp=2 and external_dp=2
+  - VLLM_USE_V1=0 torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
   - torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
-  # test with torchrun tp=2 and pp=2
+  # test with tp=2 and pp=2
   - PP_SIZE=2 torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
-  # test with torchrun tp=4 and dp=1
-  - TP_SIZE=4 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
-  # test with torchrun tp=2, pp=2 and dp=1
-  - PP_SIZE=2 TP_SIZE=2 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
-  # test with torchrun tp=1 and dp=4 with ep
-  - DP_SIZE=4 ENABLE_EP=1 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
-  # test with torchrun tp=2 and dp=2 with ep
-  - TP_SIZE=2 DP_SIZE=2 ENABLE_EP=1 torchrun --nproc-per-node=4 distributed/test_torchrun_example_moe.py
   # test with internal dp
   - python3 ../examples/offline_inference/data_parallel.py --enforce-eager
-  - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
-  - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
-  - TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/distributed/test_internal_lb_dp.py
-  - TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/distributed/test_hybrid_lb_dp.py
+  - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/test_async_llm_dp.py
+  - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/test_external_lb_dp.py
+  - TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/test_internal_lb_dp.py
+  - TP_SIZE=1 DP_SIZE=4 pytest -v -s v1/test_hybrid_lb_dp.py
   - pytest -v -s v1/engine/test_engine_core_client.py::test_kv_cache_events_dp
   - pytest -v -s distributed/test_utils.py
   - pytest -v -s compile/test_basic_correctness.py
   - pytest -v -s distributed/test_pynccl.py
   - pytest -v -s distributed/test_events.py
-  - pytest -v -s distributed/test_symm_mem_allreduce.py
   # TODO: create a dedicated test section for multi-GPU example tests
   # when we have multiple distributed example tests
   - pushd ../examples/offline_inference
@@ -205,24 +179,6 @@ steps:
   - VLLM_ALLOW_INSECURE_SERIALIZATION=1 RAY_DEDUP_LOGS=0 python3 rlhf_colocate.py
   - popd
 
-- label: Distributed Tests (8 GPUs) # 4min
-  timeout_in_minutes: 10
-  gpu: h100
-  num_gpus: 8
-  working_dir: "/vllm-workspace/tests"
-  source_file_dependencies:
-  - examples/offline_inference/torchrun_dp_example.py
-  - vllm/config/parallel.py
-  - vllm/distributed/
-  - vllm/v1/engine/llm_engine.py
-  - vllm/v1/executor/uniproc_executor.py
-  - vllm/v1/worker/gpu_worker.py
-  commands:
-  # https://github.com/NVIDIA/nccl/issues/1838
-  - export NCCL_CUMEM_HOST_ENABLE=0
-  # test with torchrun tp=2 and dp=4 with ep
-  - torchrun --nproc-per-node=8 ../examples/offline_inference/torchrun_dp_example.py --tp-size=2 --pp-size=1 --dp-size=4 --enable-ep
-
 - label: EPLB Algorithm Test # 5min
   timeout_in_minutes: 15
   working_dir: "/vllm-workspace/tests"
@@ -248,14 +204,16 @@ steps:
   num_gpus: 2
   source_file_dependencies:
   - vllm/
-  - tests/v1/tracing
+  - tests/metrics
+  - tests/tracing
   commands:
+  - pytest -v -s metrics
   - "pip install \
       'opentelemetry-sdk>=1.26.0' \
       'opentelemetry-api>=1.26.0' \
       'opentelemetry-exporter-otlp>=1.26.0' \
       'opentelemetry-semantic-conventions-ai>=0.4.1'"
-  - pytest -v -s v1/tracing
+  - pytest -v -s tracing
 
 ##### fast check tests  #####
 #####  1 GPU test  #####
@@ -316,44 +274,23 @@ steps:
     - tests/v1
   commands:
     # split the test to avoid interference
-    - pytest -v -s -m 'not cpu_test' v1/core
+    - pytest -v -s v1/core
     - pytest -v -s v1/executor
-    - pytest -v -s v1/kv_offload
     - pytest -v -s v1/sample
     - pytest -v -s v1/logits_processors
     - pytest -v -s v1/worker
+    - pytest -v -s v1/structured_output
     - pytest -v -s v1/spec_decode
-    - pytest -v -s -m 'not cpu_test' v1/kv_connector/unit
-    - pytest -v -s -m 'not cpu_test' v1/metrics
+    - pytest -v -s v1/kv_connector/unit
+    - pytest -v -s v1/metrics
+    - pytest -v -s v1/test_serial_utils.py
+    - pytest -v -s v1/test_utils.py
     - pytest -v -s v1/test_oracle.py
-    - pytest -v -s v1/test_request.py
+    - pytest -v -s v1/test_metrics_reader.py
     # Integration test for streaming correctness (requires special branch).
     - pip install -U git+https://github.com/robertgshaw2-redhat/lm-evaluation-harness.git@streaming-api
     - pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine
 
-- label: V1 Test attention (H100) # 10min
-  timeout_in_minutes: 30
-  gpu: h100
-  source_file_dependencies:
-    - vllm/v1/attention
-    - tests/v1/attention
-  commands:
-    - pytest -v -s v1/attention
-
-- label: V1 Test others (CPU) # 5 mins
-  source_file_dependencies:
-    - vllm/
-    - tests/v1
-  no_gpu: true
-  commands:
-    # split the test to avoid interference
-    - pytest -v -s -m 'cpu_test' v1/core
-    - pytest -v -s v1/structured_output
-    - pytest -v -s v1/test_serial_utils.py
-    - pytest -v -s -m 'cpu_test' v1/kv_connector/unit
-    - pytest -v -s -m 'cpu_test' v1/metrics
-
-
 - label: Examples Test # 30min
   timeout_in_minutes: 45
   mirror_hardwares: [amdexperimental]
@@ -372,14 +309,13 @@ steps:
     - python3 offline_inference/vision_language.py --seed 0
     - python3 offline_inference/vision_language_pooling.py --seed 0
     - python3 offline_inference/vision_language_multi_image.py --seed 0
-    - python3 others/tensorize_vllm_model.py --model facebook/opt-125m serialize --serialized-directory /tmp/ --suffix v1 && python3 others/tensorize_vllm_model.py --model facebook/opt-125m deserialize --path-to-tensors /tmp/vllm/facebook/opt-125m/v1/model.tensors
+    - VLLM_USE_V1=0 python3 others/tensorize_vllm_model.py --model facebook/opt-125m serialize --serialized-directory /tmp/ --suffix v1 && python3 others/tensorize_vllm_model.py --model facebook/opt-125m deserialize --path-to-tensors /tmp/vllm/facebook/opt-125m/v1/model.tensors
+    - python3 offline_inference/encoder_decoder.py
     - python3 offline_inference/encoder_decoder_multimodal.py --model-type whisper --seed 0
     - python3 offline_inference/basic/classify.py
     - python3 offline_inference/basic/embed.py
     - python3 offline_inference/basic/score.py
-    - python3 offline_inference/spec_decode.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
-    # https://github.com/vllm-project/vllm/pull/26682 uses slightly more memory in PyTorch 2.9+ causing this test to OOM in 1xL4 GPU
-    - python3 offline_inference/spec_decode.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
+    - VLLM_USE_V1=0 python3 offline_inference/profiling.py --model facebook/opt-125m run_num_steps --num-steps 2
 
 - label: Platform Tests (CUDA) # 4min
   timeout_in_minutes: 15
@@ -414,12 +350,7 @@ steps:
       --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT \
       --ignore=lora/test_chatglm3_tp.py \
       --ignore=lora/test_llama_tp.py \
-      --ignore=lora/test_llm_with_multi_loras.py \
-      --ignore=lora/test_olmoe_tp.py \
-      --ignore=lora/test_deepseekv2_tp.py \
-      --ignore=lora/test_gptoss.py \
-      --ignore=lora/test_qwen3moe_tp.py
-
+      --ignore=lora/test_llm_with_multi_loras.py
   parallelism: 4
 
 - label: PyTorch Compilation Unit Tests # 15min
@@ -433,12 +364,11 @@ steps:
     - pytest -v -s compile/test_pass_manager.py
     - pytest -v -s compile/test_fusion.py
     - pytest -v -s compile/test_fusion_attn.py
-    - pytest -v -s compile/test_functionalization.py
     - pytest -v -s compile/test_silu_mul_quant_fusion.py
+    - pytest -v -s compile/test_sequence_parallelism.py
+    - pytest -v -s compile/test_async_tp.py
     - pytest -v -s compile/test_fusion_all_reduce.py
     - pytest -v -s compile/test_decorator.py
-    - pytest -v -s compile/test_noop_elimination.py
-    - pytest -v -s compile/test_aot_compile.py
 
 - label: PyTorch Fullgraph Smoke Test # 15min
   timeout_in_minutes: 30
@@ -449,10 +379,14 @@ steps:
   - tests/compile
   commands:
   - pytest -v -s compile/test_basic_correctness.py
-  - pytest -v -s compile/piecewise/
+  # these tests need to be separated, cannot combine
+  - pytest -v -s compile/piecewise/test_simple.py
+  - pytest -v -s compile/piecewise/test_toy_llama.py
+  - pytest -v -s compile/piecewise/test_full_cudagraph.py
+  - pytest -v -s compile/piecewise/test_multiple_graphs.py
 
-- label: PyTorch Fullgraph Test # 22min
-  timeout_in_minutes: 35
+- label: PyTorch Fullgraph Test # 20min
+  timeout_in_minutes: 30
   mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
@@ -460,19 +394,6 @@ steps:
   - tests/compile
   commands:
   - pytest -v -s compile/test_full_graph.py
-  - pytest -v -s compile/test_fusions_e2e.py
-
-- label: Cudagraph test
-  timeout_in_minutes: 20
-  mirror_hardwares: [amdexperimental]
-  source_file_dependencies:
-  - tests/v1/cudagraph
-  - vllm/v1/cudagraph_dispatcher.py
-  - vllm/config/compilation.py
-  - vllm/compilation
-  commands:
-    - pytest -v -s v1/cudagraph/test_cudagraph_dispatch.py
-    - pytest -v -s v1/cudagraph/test_cudagraph_mode.py
 
 - label: Kernels Core Operation Test # 48min
   timeout_in_minutes: 75
@@ -480,9 +401,8 @@ steps:
   source_file_dependencies:
   - csrc/
   - tests/kernels/core
-  - tests/kernels/test_top_k_per_row.py
   commands:
-    - pytest -v -s kernels/core kernels/test_top_k_per_row.py
+    - pytest -v -s kernels/core
 
 - label: Kernels Attention Test %N # 23min
   timeout_in_minutes: 35
@@ -516,8 +436,6 @@ steps:
   - tests/kernels/moe
   - vllm/model_executor/layers/fused_moe/
   - vllm/distributed/device_communicators/
-  - vllm/envs.py
-  - vllm/config
   commands:
     - pytest -v -s kernels/moe --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT
   parallelism: 2
@@ -528,23 +446,33 @@ steps:
   source_file_dependencies:
   - csrc/mamba/
   - tests/kernels/mamba
-  - vllm/model_executor/layers/mamba/ops
   commands:
     - pytest -v -s kernels/mamba
 
-- label: Model Executor Test # 23min
-  timeout_in_minutes: 35
+- label: Tensorizer Test # 14min
+  timeout_in_minutes: 25
   mirror_hardwares: [amdexperimental]
   source_file_dependencies:
-  - vllm/model_executor
-  - tests/model_executor
+  - vllm/model_executor/model_loader
+  - tests/tensorizer_loader
   - tests/entrypoints/openai/test_tensorizer_entrypoint.py
   commands:
     - apt-get update && apt-get install -y curl libsodium23
     - export VLLM_WORKER_MULTIPROC_METHOD=spawn
-    - pytest -v -s model_executor
+    - pytest -v -s tensorizer_loader
     - pytest -v -s entrypoints/openai/test_tensorizer_entrypoint.py
 
+- label: Model Executor Test # 7min
+  timeout_in_minutes: 20
+  mirror_hardwares: [amdexperimental]
+  source_file_dependencies:
+  - vllm/model_executor
+  - tests/model_executor
+  commands:
+    - apt-get update && apt-get install -y curl libsodium23
+    - export VLLM_WORKER_MULTIPROC_METHOD=spawn
+    - pytest -v -s model_executor
+
 - label: Benchmarks # 11min
   timeout_in_minutes: 20
   mirror_hardwares: [amdexperimental]
@@ -573,13 +501,8 @@ steps:
   commands:
   # temporary install here since we need nightly, will move to requirements/test.in
   # after torchao 0.12 release, and pin a working version of torchao nightly here
-
-  # since torchao nightly is only compatible with torch nightly currently
-  # https://github.com/pytorch/ao/issues/2919, we'll have to skip new torchao tests for now
-  # we can only upgrade after this is resolved
-  # TODO(jerryzh168): resolve the above comment
-  - uv pip install --system torchao==0.13.0 --index-url https://download.pytorch.org/whl/cu129
-  - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
+  - pip install --pre torchao==0.13.0.dev20250814 --index-url https://download.pytorch.org/whl/nightly/cu128
+  - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization
 
 - label: LM Eval Small Models # 53min
   timeout_in_minutes: 75
@@ -600,6 +523,15 @@ steps:
   commands: # LMEval+Transcription WER check
   - pytest -s entrypoints/openai/correctness/
 
+- label: Encoder Decoder tests # 12min
+  timeout_in_minutes: 20
+  mirror_hardwares: [amdexperimental]
+  source_file_dependencies:
+  - vllm/
+  - tests/encoder_decoder
+  commands:
+    - pytest -v -s encoder_decoder
+
 - label: OpenAI-Compatible Tool Use # 23 min
   timeout_in_minutes: 35
   mirror_hardwares: [amdexperimental]
@@ -607,105 +539,43 @@ steps:
   source_file_dependencies:
     - vllm/
     - tests/tool_use
+    - tests/mistral_tool_use
   commands:
-    - pytest -v -s -m 'not cpu_test' tool_use
-
-- label: OpenAI-Compatible Tool Use (CPU) # 5 mins
-  timeout_in_minutes: 10
-  source_file_dependencies:
-    - vllm/
-    - tests/tool_use
-  no_gpu: true
-  commands:
-    - pytest -v -s -m 'cpu_test' tool_use
+    - pytest -v -s tool_use
+    - pytest -v -s mistral_tool_use
 
 #####  models test  #####
 
-- label: Basic Models Tests (Initialization)
-  timeout_in_minutes: 45
+- label: Basic Models Test # 57min
+  timeout_in_minutes: 75
   mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
   - vllm/
-  - tests/models/test_initialization.py
+  - tests/models
   commands:
-    # Run a subset of model initialization tests
-    - pytest -v -s models/test_initialization.py::test_can_initialize_small_subset
+    - pytest -v -s models/test_transformers.py
+    - pytest -v -s models/test_registry.py
+    - pytest -v -s models/test_utils.py
+    - pytest -v -s models/test_vision.py
+    - pytest -v -s models/test_initialization.py
 
-- label: Basic Models Tests (Extra Initialization) %N
+- label: Language Models Test (Standard) # 35min
   timeout_in_minutes: 45
   mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
-  - vllm/model_executor/models/
-  - tests/models/test_initialization.py
-  commands:
-    # Only when vLLM model source is modified - test initialization of a large
-    # subset of supported models (the complement of the small subset in the above
-    # test.) Also run if model initialization test file is modified
-    - pytest -v -s models/test_initialization.py \
-             -k 'not test_can_initialize_small_subset' \
-             --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT \
-             --shard-id=$$BUILDKITE_PARALLEL_JOB
-  parallelism: 2
-
-- label: Basic Models Tests (Other)
-  timeout_in_minutes: 45
-  mirror_hardwares: [amdexperimental]
-  torch_nightly: true
-  source_file_dependencies:
-  - vllm/
-  - tests/models/test_transformers.py
-  - tests/models/test_registry.py
-  commands:
-    - pytest -v -s models/test_transformers.py models/test_registry.py
-
-- label: Basic Models Test (Other CPU) # 5min
-  timeout_in_minutes: 10
-  torch_nightly: true
-  source_file_dependencies:
-  - vllm/
-  - tests/models/test_utils.py
-  - tests/models/test_vision.py
-  no_gpu: true
-  commands:
-    - pytest -v -s models/test_utils.py models/test_vision.py
-
-- label: Language Models Tests (Standard)
-  timeout_in_minutes: 25
-  mirror_hardwares: [amdexperimental]
-  torch_nightly: true
-  source_file_dependencies:
   - vllm/
   - tests/models/language
   commands:
-    # Test standard language models, excluding a subset of slow tests
     - pip freeze | grep -E 'torch'
-    - pytest -v -s models/language -m 'core_model and (not slow_test)'
+    - pytest -v -s models/language -m core_model
 
-- label: Language Models Tests (Extra Standard) %N
+- label: Language Models Test (Hybrid) # 35 min
   timeout_in_minutes: 45
   mirror_hardwares: [amdexperimental]
   torch_nightly: true
   source_file_dependencies:
-  - vllm/model_executor/models/
-  - tests/models/language/pooling/test_embedding.py
-  - tests/models/language/generation/test_common.py
-  - tests/models/language/pooling/test_classification.py
-  commands:
-    # Shard slow subset of standard language models tests. Only run when model
-    # source is modified, or when specified test files are modified
-    - pip freeze | grep -E 'torch'
-    - pytest -v -s models/language -m 'core_model and slow_test' \
-             --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT \
-             --shard-id=$$BUILDKITE_PARALLEL_JOB
-  parallelism: 2
-
-- label: Language Models Tests (Hybrid) %N
-  timeout_in_minutes: 75
-  mirror_hardwares: [amdexperimental]
-  torch_nightly: true
-  source_file_dependencies:
   - vllm/
   - tests/models/language/generation
   commands:
@@ -713,12 +583,7 @@ steps:
     # Note: also needed to run plamo2 model in vLLM
     - uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.2.5'
     - uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.2'
-    # Shard hybrid language model tests
-    - pytest -v -s models/language/generation \
-                   -m hybrid_model \
-                   --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT \
-                   --shard-id=$$BUILDKITE_PARALLEL_JOB
-  parallelism: 2
+    - pytest -v -s models/language/generation -m hybrid_model
 
 - label: Language Models Test (Extended Generation) # 80min
   timeout_in_minutes: 110
@@ -728,22 +593,10 @@ steps:
   - vllm/
   - tests/models/language/generation
   commands:
-    # Install fast path packages for testing against transformers
-    # Note: also needed to run plamo2 model in vLLM
-    - uv pip install --system --no-build-isolation 'git+https://github.com/state-spaces/mamba@v2.2.5'
-    - uv pip install --system --no-build-isolation 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.2'
+    # Install causal-conv1d for plamo2 models here, as it is not compatible with pip-compile.
+    - pip install 'git+https://github.com/Dao-AILab/causal-conv1d@v1.5.0.post8'
     - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
 
-- label: Language Models Test (PPL)
-  timeout_in_minutes: 110
-  mirror_hardwares: [amdexperimental]
-  optional: true
-  source_file_dependencies:
-  - vllm/
-  - tests/models/language/generation_ppl_test
-  commands:
-    - pytest -v -s models/language/generation_ppl_test
-
 - label: Language Models Test (Extended Pooling)  # 36min
   timeout_in_minutes: 50
   mirror_hardwares: [amdexperimental]
@@ -754,16 +607,6 @@ steps:
   commands:
     - pytest -v -s models/language/pooling -m 'not core_model'
 
-- label: Language Models Test (MTEB)
-  timeout_in_minutes: 110
-  mirror_hardwares: [amdexperimental]
-  optional: true
-  source_file_dependencies:
-  - vllm/
-  - tests/models/language/pooling_mteb_test
-  commands:
-    - pytest -v -s models/language/pooling_mteb_test
-
 - label: Multi-Modal Processor Test # 44min
   timeout_in_minutes: 60
   source_file_dependencies:
@@ -784,17 +627,7 @@ steps:
     - pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
     - pip freeze | grep -E 'torch'
     - pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/processing
-    - cd .. && VLLM_WORKER_MULTIPROC_METHOD=spawn pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model  # Otherwise, mp_method="spawn" doesn't work
-
-- label: Multi-Modal Accuracy Eval (Small Models) # 50min
-  timeout_in_minutes: 70
-  working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
-  source_file_dependencies:
-  - vllm/multimodal/
-  - vllm/inputs/
-  - vllm/v1/core/
-  commands:
-  - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-mm-small.txt --tp-size=1
+    - cd .. && pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model  # Otherwise, mp_method="spawn" doesn't work
 
 - label: Multi-Modal Models Test (Extended) 1
   mirror_hardwares: [amdexperimental]
@@ -851,16 +684,14 @@ steps:
   commands:
     - pip install --upgrade git+https://github.com/huggingface/transformers
     - pytest -v -s tests/models/test_initialization.py
-    - pytest -v -s tests/models/test_transformers.py
     - pytest -v -s tests/models/multimodal/processing/
     - pytest -v -s tests/models/multimodal/test_mapping.py
     - python3 examples/offline_inference/basic/chat.py
+    - python3 examples/offline_inference/audio_language.py --model-type whisper
     - python3 examples/offline_inference/vision_language.py --model-type qwen2_5_vl
-    # Whisper needs spawn method to avoid deadlock
-    - VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/offline_inference/audio_language.py --model-type whisper
 
-- label: Blackwell Test # 21 min
-  timeout_in_minutes: 30
+- label: Blackwell Test # 38 min
+  timeout_in_minutes: 60
   working_dir: "/vllm-workspace/"
   gpu: b200
   # optional: true
@@ -873,6 +704,8 @@ steps:
   - vllm/model_executor/layers/fused_moe/flashinfer_cutlass_prepare_finalize.py
   - vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
   - vllm/v1/attention/backends/flashinfer.py
+  - vllm/compilation/fusion.py
+  - vllm/compilation/fusion_attn.py
   commands:
     - nvidia-smi
     - python3 examples/offline_inference/basic/chat.py
@@ -880,82 +713,21 @@ steps:
     # num_heads2 broken by https://github.com/flashinfer-ai/flashinfer/issues/1353
     - pytest -v -s tests/kernels/attention/test_flashinfer.py -k 'not num_heads2'
     - pytest -v -s tests/kernels/attention/test_flashinfer_trtllm_attention.py
-    - pytest -v -s tests/kernels/attention/test_cutlass_mla_decode.py
-    - pytest -v -s tests/kernels/attention/test_flashinfer_mla_decode.py
+    - pytest -v -s tests/kernels/test_cutlass_mla_decode.py
     # Quantization
     - pytest -v -s tests/kernels/quantization/test_cutlass_scaled_mm.py -k 'fp8'
     - pytest -v -s tests/kernels/quantization/test_nvfp4_quant.py
-    - pytest -v -s tests/kernels/quantization/test_silu_mul_nvfp4_quant.py
+    - pytest -v -s tests/kernels/quantization/test_silu_nvfp4_quant_fusion.py
     - pytest -v -s tests/kernels/quantization/test_nvfp4_scaled_mm.py
     - pytest -v -s tests/kernels/quantization/test_flashinfer_scaled_mm.py
     - pytest -v -s tests/kernels/quantization/test_flashinfer_nvfp4_scaled_mm.py
-    - pytest -v -s tests/kernels/quantization/test_nvfp4_qutlass.py
-    - pytest -v -s tests/kernels/quantization/test_mxfp4_qutlass.py
     - pytest -v -s tests/kernels/moe/test_nvfp4_moe.py
-    - pytest -v -s tests/kernels/moe/test_ocp_mx_moe.py
-    - pytest -v -s tests/kernels/moe/test_flashinfer.py
-
-- label: Blackwell Fusion Tests # 30 min
-  timeout_in_minutes: 40
-  working_dir: "/vllm-workspace/"
-  gpu: b200
-  source_file_dependencies:
-  - csrc/quantization/fp4/
-  - vllm/model_executor/layers/quantization/utils/flashinfer_utils.py
-  - vllm/v1/attention/backends/flashinfer.py
-  - vllm/compilation/
-  # can affect pattern matching
-  - vllm/model_executor/layers/layernorm.py
-  - vllm/model_executor/layers/activation.py
-  - vllm/model_executor/layers/quantization/input_quant_fp8.py
-  commands:
-    - nvidia-smi
-    - pytest -v -s tests/compile/test_fusion_attn.py
-    - pytest -v -s tests/compile/test_silu_mul_quant_fusion.py
-    # this runner has 2 GPUs available even though num_gpus=2 is not set
+    - pytest -v -s tests/kernels/moe/test_mxfp4_moe.py
+    # Fusion
     - pytest -v -s tests/compile/test_fusion_all_reduce.py
-    - pytest -v -s tests/compile/test_fusions_e2e.py
-
-- label: Blackwell GPT-OSS Eval
-  timeout_in_minutes: 60
-  working_dir: "/vllm-workspace/"
-  gpu: b200
-  optional: true # run on nightlies
-  source_file_dependencies:
-  - tests/evals/gpt_oss
-  - vllm/model_executor/models/gpt_oss.py
-  - vllm/model_executor/layers/quantization/mxfp4.py
-  - vllm/v1/attention/backends/flashinfer.py
-  commands:
-    - uv pip install --system 'gpt-oss[eval]==0.0.5'
-    - pytest -s -v tests/evals/gpt_oss/test_gpqa_correctness.py --model openai/gpt-oss-20b --metric 0.58
-
-- label: Blackwell Quantized MoE Test
-  timeout_in_minutes: 60
-  working_dir: "/vllm-workspace/"
-  gpu: b200
-  source_file_dependencies:
-  - tests/quantization/test_blackwell_moe.py
-  - vllm/model_executor/models/deepseek_v2.py
-  - vllm/model_executor/models/gpt_oss.py
-  - vllm/model_executor/models/llama4.py
-  - vllm/model_executor/layers/fused_moe
-  - vllm/model_executor/layers/quantization/compressed_tensors
-  - vllm/model_executor/layers/quantization/modelopt.py
-  - vllm/model_executor/layers/quantization/mxfp4.py
-  - vllm/v1/attention/backends/flashinfer.py
-  commands:
-    - pytest -s -v tests/quantization/test_blackwell_moe.py
-
-- label: Blackwell LM Eval Small Models
-  timeout_in_minutes: 120
-  gpu: b200
-  optional: true # run on nightlies
-  source_file_dependencies:
-  - csrc/
-  - vllm/model_executor/layers/quantization
-  commands:
-  - pytest -s -v evals/gsm8k/test_gsm8k_correctness.py --config-list-file=configs/models-blackwell.txt --tp-size=1
+    - pytest -v -s tests/compile/test_fusion_attn.py::test_attention_quant_pattern
+    - pytest -v -s tests/kernels/moe/test_flashinfer.py
+    - pytest -v -s tests/compile/test_silu_mul_quant_fusion.py
 
 #####  1 GPU test  #####
 #####  multi gpus test  #####
@@ -971,8 +743,6 @@ steps:
   commands:
   - pytest -v -s distributed/test_comm_ops.py
   - pytest -v -s distributed/test_shm_broadcast.py
-  - pytest -v -s distributed/test_shm_buffer.py
-  - pytest -v -s distributed/test_shm_storage.py
 
 - label: 2 Node Tests (4 GPUs in total) # 16min
   timeout_in_minutes: 30
@@ -999,61 +769,46 @@ steps:
     - NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed'
     - python3 ../examples/offline_inference/data_parallel.py --dp-size=2 --tp-size=1 --node-size=2 --node-rank=1 --master-addr=192.168.10.10 --master-port=12345 --enforce-eager --trust-remote-code
 
-- label: Distributed Tests (2 GPUs) # 68min
-  timeout_in_minutes: 90
+- label: Distributed Tests (2 GPUs) # 110min
+  timeout_in_minutes: 150
   mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
   num_gpus: 2
   source_file_dependencies:
-  - vllm/compilation/
   - vllm/distributed/
   - vllm/engine/
   - vllm/executor/
-  - vllm/worker/worker_base.py
-  - vllm/v1/engine/
-  - vllm/v1/worker/
-  - tests/compile/test_basic_correctness.py
-  - tests/compile/test_wrapper.py
+  - vllm/model_executor/models/
   - tests/distributed/
-  - tests/entrypoints/llm/test_collective_rpc.py
-  - tests/v1/distributed
+  - vllm/compilation
+  - vllm/worker/worker_base.py
+  - vllm/worker/worker.py
+  - vllm/worker/model_runner.py
+  - entrypoints/llm/test_collective_rpc.py
+  - tests/v1/test_async_llm_dp.py
+  - tests/v1/test_external_lb_dp.py
   - tests/v1/entrypoints/openai/test_multi_api_servers.py
-  - tests/v1/shutdown
-  - tests/v1/worker/test_worker_memory_snapshot.py
+  - vllm/v1/engine/
   commands:
-  # https://github.com/NVIDIA/nccl/issues/1838
-  - export NCCL_CUMEM_HOST_ENABLE=0
-  - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_async_llm_dp.py
-  - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/distributed/test_external_lb_dp.py
+  - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/test_async_llm_dp.py
+  - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/test_external_lb_dp.py
   - DP_SIZE=2 pytest -v -s v1/entrypoints/openai/test_multi_api_servers.py
   - pytest -v -s entrypoints/llm/test_collective_rpc.py
   - pytest -v -s ./compile/test_basic_correctness.py
   - pytest -v -s ./compile/test_wrapper.py
   - VLLM_TEST_SAME_HOST=1 torchrun --nproc-per-node=4 distributed/test_same_node.py | grep 'Same node test passed'
-  - VLLM_TEST_SAME_HOST=1 VLLM_TEST_WITH_DEFAULT_DEVICE_SET=1 torchrun --nproc-per-node=4 distributed/test_same_node.py | grep 'Same node test passed'
-  - pytest -v -s distributed/test_sequence_parallel.py
-  - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s v1/shutdown
-  - pytest -v -s v1/worker/test_worker_memory_snapshot.py
-
-- label: Distributed Model Tests (2 GPUs) # 37min
-  timeout_in_minutes: 50
-  mirror_hardwares: [amdexperimental]
-  working_dir: "/vllm-workspace/tests"
-  num_gpus: 2
-  source_file_dependencies:
-  - vllm/model_executor/model_loader/sharded_state_loader.py
-  - vllm/model_executor/models/
-  - tests/basic_correctness/
-  - tests/model_executor/model_loader/test_sharded_state_loader.py
-  - tests/models/
-  commands:
   - TARGET_TEST_SUITE=L4 pytest basic_correctness/ -v -s -m 'distributed(num_gpus=2)'
-  - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s model_executor/model_loader/test_sharded_state_loader.py
   # Avoid importing model tests that cause CUDA reinitialization error
   - pytest models/test_transformers.py -v -s -m 'distributed(num_gpus=2)'
   - pytest models/language -v -s -m 'distributed(num_gpus=2)'
-  - pytest models/multimodal -v -s -m 'distributed(num_gpus=2)' --ignore models/multimodal/generation/test_whisper.py
-  - VLLM_WORKER_MULTIPROC_METHOD=spawn pytest models/multimodal/generation/test_whisper.py -v -s -m 'distributed(num_gpus=2)'
+  - pytest models/multimodal -v -s -m 'distributed(num_gpus=2)'
+  # test sequence parallel
+  - pytest -v -s distributed/test_sequence_parallel.py
+  # this test fails consistently.
+  # TODO: investigate and fix
+  - VLLM_USE_V1=0 CUDA_VISIBLE_DEVICES=0,1 pytest -v -s test_sharded_state_loader.py
+  - CUDA_VISIBLE_DEVICES=0,1 pytest -v -s v1/shutdown
+  - pytest -v -s models/multimodal/generation/test_maverick.py
 
 - label: Plugin Tests (2 GPUs) # 40min
   timeout_in_minutes: 60
@@ -1072,13 +827,8 @@ steps:
   # begin io_processor plugins test, all the code in between uses the prithvi_io_processor plugin
   - pip install -e ./plugins/prithvi_io_processor_plugin
   - pytest -v -s plugins_tests/test_io_processor_plugins.py
-  - pip uninstall prithvi_io_processor_plugin -y
+  - pip uninstall prithvi_io_processor_plugin -y 
   # end io_processor plugins test
-  # begin stat_logger plugins test
-  - pip install -e ./plugins/vllm_add_dummy_stat_logger
-  - pytest -v -s plugins_tests/test_stats_logger_plugins.py
-  - pip uninstall dummy_stat_logger -y
-  # end stat_logger plugins test
   # other tests continue here:
   - pytest -v -s plugins_tests/test_scheduler_plugins.py
   - pip install -e ./plugins/vllm_add_dummy_model
@@ -1101,6 +851,7 @@ steps:
   commands:
   - pytest -v -s distributed/test_pp_cudagraph.py
   - pytest -v -s distributed/test_pipeline_parallel.py
+  # - pytest -v -s distributed/test_context_parallel.py # TODO: enable it on Hopper runners or add triton MLA support
 
 - label: LoRA TP Test (Distributed) # 17 min
   timeout_in_minutes: 30
@@ -1118,14 +869,13 @@ steps:
     - pytest -v -s -x lora/test_chatglm3_tp.py
     - pytest -v -s -x lora/test_llama_tp.py
     - pytest -v -s -x lora/test_llm_with_multi_loras.py
-    - pytest -v -s -x lora/test_olmoe_tp.py
 
 
 - label: Weight Loading Multiple GPU Test  # 33min
   timeout_in_minutes: 45
   mirror_hardwares: [amdexperimental]
   working_dir: "/vllm-workspace/tests"
-  num_gpus: 2
+  num_gpus: 2 
   optional: true
   source_file_dependencies:
   - vllm/
@@ -1145,17 +895,6 @@ steps:
   commands:
     - bash weight_loading/run_model_weight_loading_test.sh -c weight_loading/models-large.txt
 
-- label: NixlConnector PD accuracy tests (Distributed) # 30min
-  timeout_in_minutes: 30
-  working_dir: "/vllm-workspace/tests"
-  num_gpus: 4
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - bash v1/kv_connector/nixl_integration/tp_config_sweep_accuracy_test.sh
-
 
 ##### multi gpus test #####
 ##### A100 test #####
@@ -1186,51 +925,9 @@ steps:
   - export VLLM_WORKER_MULTIPROC_METHOD=spawn
   - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large.txt --tp-size=4
 
-##### H100 test #####
-- label: LM Eval Large Models (H100) # optional
-  gpu: h100
-  optional: true
-  num_gpus: 4
-  working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
-  source_file_dependencies:
-  - csrc/
-  - vllm/model_executor/layers/quantization
-  commands:
-    - export VLLM_USE_DEEP_GEMM=0  # We found Triton is faster than DeepGEMM for H100
-    - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-large-hopper.txt --tp-size=4
-
-##### H200 test #####
-- label: Distributed Tests (H200) # optional
+- label: Qwen MoE EP Test # optional
   gpu: h200
   optional: true
-  working_dir: "/vllm-workspace/"
   num_gpus: 2
   commands:
-    - pytest -v -s tests/compile/test_async_tp.py
-    - pytest -v -s tests/compile/test_sequence_parallelism.py
-    - pytest -v -s tests/compile/test_fusion_all_reduce.py
-    - pytest -v -s tests/compile/test_fusions_e2e.py::test_tp2_attn_quant_allreduce_rmsnorm
-    - pytest -v -s tests/distributed/test_context_parallel.py
-    - CUDA_VISIBLE_DEVICES=1,2 VLLM_ALL2ALL_BACKEND=deepep_high_throughput VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model Qwen/Qwen1.5-MoE-A2.7B --tp-size=1  --dp-size=2 --max-model-len 2048
-
-##### B200 test #####
-- label: Distributed Tests (B200) # optional
-  gpu: b200
-  optional: true
-  working_dir: "/vllm-workspace/"
-  num_gpus: 2
-  commands:
-    - pytest -v -s tests/distributed/test_context_parallel.py
-    - pytest -v -s tests/distributed/test_nccl_symm_mem_allreduce.py
-
-##### RL Integration Tests #####
-- label: Prime-RL Integration Test # 15min
-  timeout_in_minutes: 30
-  optional: true
-  num_gpus: 2
-  working_dir: "/vllm-workspace"
-  source_file_dependencies:
-  - vllm/
-  - .buildkite/scripts/run-prime-rl-test.sh
-  commands:
-    - bash .buildkite/scripts/run-prime-rl-test.sh
+    - CUDA_VISIBLE_DEVICES=1,2 VLLM_ALL2ALL_BACKEND=deepep_high_throughput VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 /vllm-workspace/examples/offline_inference/data_parallel.py --model Qwen/Qwen1.5-MoE-A2.7B --tp-size=1  --dp-size=2 --max-model-len 2048

.coveragerc

@@ -1,47 +0,0 @@
-[run]
-# Track the installed vllm package (this is what actually gets imported during tests)
-# Use wildcard pattern to match the installed location
-source =
-    vllm
-    */dist-packages/vllm
-    */site-packages/vllm
-omit =
-    */tests/*
-    */test_*
-    */__pycache__/*
-    */build/*
-    */dist/*
-    */vllm.egg-info/*
-    */third_party/*
-    */examples/*
-    */benchmarks/*
-    */docs/*
-
-[paths]
-# Map all possible vllm locations to a canonical "vllm" path
-# This ensures coverage.combine properly merges data from different test runs
-source =
-    vllm
-    /vllm-workspace/src/vllm
-    /vllm-workspace/vllm
-    */site-packages/vllm
-    */dist-packages/vllm
-
-[report]
-exclude_lines =
-    pragma: no cover
-    def __repr__
-    if self.debug:
-    if settings.DEBUG
-    raise AssertionError
-    raise NotImplementedError
-    if 0:
-    if __name__ == .__main__.:
-    class .*\bProtocol\):
-    @(abc\.)?abstractmethod
-
-[html]
-directory = htmlcov
-
-[xml]
-output = coverage.xml

.git-blame-ignore-revs

@@ -1,4 +0,0 @@
-# Migrate from `yapf` & `isort` to `ruff`
-d6953beb91da4e9c99be4c0a1304a2d24189535c
-# Convert `Optional[x]` to `x | None` and `Union[x, y]` to `x | y`
-8fcaaf6a165e661f63fc51be906bc05b0767332f

.github/.bc-linter.yml

@@ -1,24 +0,0 @@
-# doc: https://github.com/pytorch/test-infra/blob/main/tools/stronghold/docs/bc_linter_config.md
-version: 1
-paths:
-# We temporarily disable globally, and will only enable with `annotations.include`
-# include:
-#   - "vllm/v1/attetion/*.py"
-#   - "vllm/v1/core/*.py"
-exclude:
-  - "**/*.py"
-
-scan:
-  functions: true        # check free functions and methods
-  classes: true          # check classes/dataclasses
-  public_only: true      # ignore names starting with "_" at any level
-
-annotations:
-  include:               # decorators that force‑include a symbol
-    - name: "bc_linter_include"  # matched by simple name or dotted suffix
-      propagate_to_members: false # for classes, include methods/inner classes
-  exclude:               # decorators that force‑exclude a symbol
-    - name: "bc_linter_skip"     # matched by simple name or dotted suffix
-      propagate_to_members: true  # for classes, exclude methods/inner classes
-
-excluded_violations: []  # e.g. ["ParameterRenamed", "FieldTypeChanged"]

.github/CODEOWNERS

@@ -2,86 +2,67 @@
 # for more info about CODEOWNERS file
 
 # This lists cover the "core" components of vLLM that require careful review
-/vllm/attention @LucasWilkinson
 /vllm/attention/backends/abstract.py @WoosukKwon @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
+/vllm/core @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
+/vllm/engine/llm_engine.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
 /vllm/executor/executor_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill @22quinn
-/vllm/model_executor/layers/fused_moe @mgoin @pavanimajety
-/vllm/model_executor/layers/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth @yewentao256 @pavanimajety
+/vllm/worker/worker_base.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill @22quinn
+/vllm/worker/worker.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
+/vllm/model_executor/layers/sampler.py @zhuohan123 @youkaichao @alexm-redhat @comaniac @njhill
+/vllm/model_executor/layers/quantization @mgoin @robertgshaw2-redhat @tlrmchlsmth @yewentao256
 /vllm/model_executor/layers/mamba @tdoublep
 /vllm/model_executor/model_loader @22quinn
-/vllm/multimodal @DarkLight1337 @ywang96 @NickLucche
+/vllm/multimodal @DarkLight1337 @ywang96
+/vllm/v1/sample @22quinn @houseroad
 /vllm/vllm_flash_attn @LucasWilkinson
 /vllm/lora @jeejeelee
-/vllm/reasoning @aarnphm @chaunceyjiang
-/vllm/entrypoints @aarnphm @chaunceyjiang
+/vllm/reasoning @aarnphm
+/vllm/entrypoints @aarnphm
 /vllm/compilation @zou3519 @youkaichao @ProExpertProg
-/vllm/distributed/kv_transfer @NickLucche @ApostaC
 CMakeLists.txt @tlrmchlsmth @LucasWilkinson
 
 # Any change to the VllmConfig changes can have a large user-facing impact,
 # so spam a lot of people
 /vllm/config @simon-mo @WoosukKwon @youkaichao @robertgshaw2-redhat @mgoin @tlrmchlsmth @houseroad @hmellor @yewentao256 @ProExpertProg
-/vllm/config/cache.py @simon-mo @WoosukKwon @youkaichao @robertgshaw2-redhat @mgoin @tlrmchlsmth @houseroad @hmellor @yewentao256 @ProExpertProg @heheda12345
 
 # vLLM V1
-/vllm/v1/attention @LucasWilkinson
-/vllm/v1/attention/backends/mla @pavanimajety
-/vllm/v1/attention/backends/flashinfer.py @mgoin @pavanimajety
-/vllm/v1/attention/backends/triton_attn.py @tdoublep
-/vllm/v1/core @WoosukKwon @robertgshaw2-redhat @njhill @ywang96 @comaniac @alexm-redhat @heheda12345 @ApostaC
-/vllm/v1/sample @22quinn @houseroad @njhill
-/vllm/v1/spec_decode @benchislett @luccafong
+/vllm/v1 @WoosukKwon @robertgshaw2-redhat @njhill @ywang96 @comaniac @alexm-redhat
 /vllm/v1/structured_output @mgoin @russellb @aarnphm @benchislett
-/vllm/v1/kv_cache_interface.py @heheda12345
-/vllm/v1/offloading @ApostaC
+/vllm/v1/spec_decode @benchislett @luccafong
+/vllm/v1/attention/backends/triton_attn.py @tdoublep
 
 # Test ownership
 /.buildkite/lm-eval-harness @mgoin @simon-mo
+/tests/async_engine @njhill @robertgshaw2-redhat @simon-mo
 /tests/distributed/test_multi_node_assignment.py @youkaichao
 /tests/distributed/test_pipeline_parallel.py @youkaichao
 /tests/distributed/test_same_node.py @youkaichao
-/tests/entrypoints @DarkLight1337 @robertgshaw2-redhat @simon-mo @aarnphm @NickLucche
-/tests/evals @mgoin
-/tests/kernels @mgoin @tlrmchlsmth @WoosukKwon @yewentao256
+/tests/entrypoints @DarkLight1337 @robertgshaw2-redhat @simon-mo @aarnphm
+/tests/kernels @tlrmchlsmth @WoosukKwon @yewentao256
 /tests/models @DarkLight1337 @ywang96
-/tests/multimodal @DarkLight1337 @ywang96 @NickLucche
-/tests/quantization @mgoin @robertgshaw2-redhat @yewentao256 @pavanimajety
+/tests/multimodal @DarkLight1337 @ywang96
+/tests/prefix_caching @comaniac @KuntaiDu
+/tests/quantization @mgoin @robertgshaw2-redhat @yewentao256
 /tests/test_inputs.py @DarkLight1337 @ywang96
 /tests/v1/entrypoints/llm/test_struct_output_generate.py @mgoin @russellb @aarnphm
 /tests/v1/structured_output @mgoin @russellb @aarnphm
-/tests/v1/core @WoosukKwon @robertgshaw2-redhat @njhill @ywang96 @comaniac @alexm-redhat @heheda12345 @ApostaC
 /tests/weight_loading @mgoin @youkaichao @yewentao256
 /tests/lora @jeejeelee
 /tests/models/language/generation/test_hybrid.py @tdoublep
-/tests/v1/kv_connector/nixl_integration @NickLucche
-/tests/v1/kv_connector @ApostaC
-/tests/v1/offloading @ApostaC
-
-# Transformers backend
-/vllm/model_executor/models/transformers @hmellor
-/tests/models/test_transformers.py @hmellor
 
 # Docs
-/docs/mkdocs @hmellor
-/docs/**/*.yml @hmellor
-/requirements/docs.txt @hmellor
-.readthedocs.yaml @hmellor
+/docs @hmellor
 mkdocs.yaml @hmellor
 
-# Linting
-.markdownlint.yaml @hmellor
-.pre-commit-config.yaml @hmellor
-/tools/pre_commit @hmellor
-
 # CPU
-/vllm/v1/worker/cpu* @bigPYJ1151
+/vllm/v1/worker/^cpu @bigPYJ1151
 /csrc/cpu @bigPYJ1151
 /vllm/platforms/cpu.py @bigPYJ1151
 /cmake/cpu_extension.cmake @bigPYJ1151
 /docker/Dockerfile.cpu @bigPYJ1151
 
 # Intel GPU
-/vllm/v1/worker/xpu* @jikunshang
+/vllm/v1/worker/^xpu @jikunshang
 /vllm/platforms/xpu.py @jikunshang
 /docker/Dockerfile.xpu @jikunshang
 
@@ -110,20 +91,3 @@ mkdocs.yaml @hmellor
 /vllm/v1/attention/backends/mla/rocm*.py @gshtras
 /vllm/attention/ops/rocm*.py @gshtras
 /vllm/model_executor/layers/fused_moe/rocm*.py @gshtras
-
-# TPU
-/vllm/v1/worker/tpu* @NickLucche
-/vllm/platforms/tpu.py @NickLucche
-/vllm/v1/sample/tpu @NickLucche
-/vllm/tests/v1/tpu @NickLucche
-
-# KVConnector installation files
-/requirements/kv_connectors.txt @NickLucche
-
-# Pooling models
-/examples/*/pooling/ @noooop
-/tests/models/*/pooling* @noooop
-/tests/entrypoints/pooling @noooop
-/vllm/config/pooler.py @noooop
-/vllm/pooling_params.py @noooop
-/vllm/model_executor/layers/pooler.py @noooop

.github/ISSUE_TEMPLATE/750-RFC.yml

@@ -43,6 +43,10 @@ body:
       Any other things you would like to mention.
   validations:
     required: false
+- type: markdown
+  attributes:
+    value: >
+      Thanks for contributing 🎉! The vLLM core team hosts a biweekly RFC review session at 9:30AM Pacific Time, while most RFCs can be discussed online, you can optionally sign up for a slot to discuss your RFC online [here](https://docs.google.com/document/d/1CiLVBZeIVfR7_PNAKVSusxpceywkoOOB78qoWqHvSZc/edit).
 - type: checkboxes
   id: askllm
   attributes:

.github/mergify.yml

@@ -2,7 +2,6 @@ pull_request_rules:
 - name: label-documentation
   description: Automatically apply documentation label
   conditions:
-    - label != stale
     - or:
       - files~=^[^/]+\.md$
       - files~=^docs/
@@ -11,13 +10,10 @@ pull_request_rules:
     label:
       add:
         - documentation
-    comment:
-      message: "Documentation preview: https://vllm--{{number}}.org.readthedocs.build/en/{{number}}/"
 
 - name: label-ci-build
   description: Automatically apply ci/build label
   conditions:
-    - label != stale
     - or:
       - files~=^\.github/
       - files~=\.buildkite/
@@ -34,7 +30,6 @@ pull_request_rules:
 - name: label-deepseek
   description: Automatically apply deepseek label
   conditions:
-    - label != stale
     - or:
       - files~=^examples/.*deepseek.*\.py
       - files~=^tests/.*deepseek.*\.py
@@ -51,7 +46,6 @@ pull_request_rules:
 - name: label-frontend
   description: Automatically apply frontend label
   conditions:
-    - label != stale
     - files~=^vllm/entrypoints/
   actions:
     label:
@@ -61,7 +55,6 @@ pull_request_rules:
 - name: label-llama
   description: Automatically apply llama label
   conditions:
-    - label != stale
     - or:
       - files~=^examples/.*llama.*\.py
       - files~=^tests/.*llama.*\.py
@@ -77,7 +70,6 @@ pull_request_rules:
 - name: label-multi-modality
   description: Automatically apply multi-modality label
   conditions:
-    - label != stale
     - or:
       - files~=^vllm/multimodal/
       - files~=^tests/multimodal/
@@ -91,7 +83,6 @@ pull_request_rules:
 - name: label-new-model
   description: Automatically apply new-model label
   conditions:
-    - label != stale
     - and:
       - files~=^vllm/model_executor/models/
       - files=vllm/model_executor/models/registry.py
@@ -103,12 +94,11 @@ pull_request_rules:
 - name: label-performance
   description: Automatically apply performance label
   conditions:
-    - label != stale
     - or:
       - files~=^benchmarks/
       - files~=^vllm/benchmarks/
       - files~=^tests/benchmarks/
-      - files~=^\.buildkite/performance-benchmarks/
+      - files~=^\.buildkite/nightly-benchmarks/
   actions:
     label:
       add:
@@ -117,7 +107,6 @@ pull_request_rules:
 - name: label-qwen
   description: Automatically apply qwen label
   conditions:
-    - label != stale
     - or:
       - files~=^examples/.*qwen.*\.py
       - files~=^tests/.*qwen.*\.py
@@ -132,20 +121,12 @@ pull_request_rules:
 - name: label-gpt-oss
   description: Automatically apply gpt-oss label
   conditions:
-    - label != stale
     - or:
       - files~=^examples/.*gpt[-_]?oss.*\.py
       - files~=^tests/.*gpt[-_]?oss.*\.py
-      - files~=^tests/entrypoints/openai/test_response_api_with_harmony.py
-      - files~=^tests/entrypoints/test_context.py
       - files~=^vllm/model_executor/models/.*gpt[-_]?oss.*\.py
       - files~=^vllm/model_executor/layers/.*gpt[-_]?oss.*\.py
-      - files~=^vllm/entrypoints/harmony_utils.py
-      - files~=^vllm/entrypoints/tool_server.py
-      - files~=^vllm/entrypoints/tool.py
-      - files~=^vllm/entrypoints/context.py
       - title~=(?i)gpt[-_]?oss
-      - title~=(?i)harmony
   actions:
     label:
       add:
@@ -154,7 +135,6 @@ pull_request_rules:
 - name: label-rocm
   description: Automatically apply rocm label
   conditions:
-    - label != stale
     - or:
       - files~=^csrc/rocm/
       - files~=^docker/Dockerfile.rocm
@@ -175,7 +155,6 @@ pull_request_rules:
 - name: label-structured-output
   description: Automatically apply structured-output label
   conditions:
-    - label != stale
     - or:
       - files~=^benchmarks/structured_schemas/
       - files=benchmarks/benchmark_serving_structured_output.py
@@ -185,7 +164,7 @@ pull_request_rules:
       - files=examples/online_serving/openai_chat_completion_structured_outputs.py
       - files=examples/online_serving/openai_chat_completion_structured_outputs_with_reasoning.py
       - files~=^tests/v1/structured_output/
-      - files=tests/v1/entrypoints/llm/test_struct_output_generate.py
+      - files=tests/v1/entrypoints/llm/test_guided_generate.py
       - files~=^vllm/v1/structured_output/
   actions:
     label:
@@ -195,7 +174,6 @@ pull_request_rules:
 - name: label-speculative-decoding
   description: Automatically apply speculative-decoding label
   conditions:
-    - label != stale
     - or:
       - files~=^vllm/v1/spec_decode/
       - files~=^tests/v1/spec_decode/
@@ -211,7 +189,6 @@ pull_request_rules:
 - name: label-v1
   description: Automatically apply v1 label
   conditions:
-    - label != stale
     - or:
       - files~=^vllm/v1/
       - files~=^tests/v1/
@@ -224,7 +201,6 @@ pull_request_rules:
   description: Automatically apply tpu label
   # Keep this list in sync with `label-tpu-remove` conditions
   conditions:
-    - label != stale
     - or:
       - files~=tpu.py
       - files~=_tpu
@@ -240,7 +216,6 @@ pull_request_rules:
   description: Automatically remove tpu label
   # Keep this list in sync with `label-tpu` conditions
   conditions:
-    - label != stale
     - and:
       - -files~=tpu.py
       - -files~=_tpu
@@ -255,9 +230,9 @@ pull_request_rules:
 - name: label-tool-calling
   description: Automatically add tool-calling label
   conditions:
-    - label != stale
     - or:
       - files~=^tests/tool_use/
+      - files~=^tests/mistral_tool_use/
       - files~=^tests/entrypoints/openai/tool_parsers/
       - files=tests/entrypoints/openai/test_chat_with_tool_reasoning.py
       - files~=^vllm/entrypoints/openai/tool_parsers/
@@ -274,9 +249,8 @@ pull_request_rules:
 
 - name: ping author on conflicts and add 'needs-rebase' label
   conditions:
-    - label != stale
-    - conflict
-    - -closed
+      - conflict
+      - -closed
   actions:
     label:
       add:
@@ -290,12 +264,10 @@ pull_request_rules:
 
 - name: assign reviewer for tensorizer changes
   conditions:
-    - label != stale
-    - or:
       - files~=^vllm/model_executor/model_loader/tensorizer.py
       - files~=^vllm/model_executor/model_loader/tensorizer_loader.py
       - files~=^tests/entrypoints/openai/test_tensorizer_entrypoint.py
-      - files~=^tests/model_executor/model_loader/tensorizer_loader/
+      - files~=^tests/tensorizer_loader/
   actions:
     assign:
       users:
@@ -303,7 +275,6 @@ pull_request_rules:
 
 - name: assign reviewer for modelopt changes
   conditions:
-    - label != stale
     - or:
         - files~=^vllm/model_executor/layers/quantization/modelopt\.py$
         - files~=^vllm/model_executor/layers/quantization/__init__\.py$
@@ -318,27 +289,9 @@ pull_request_rules:
 
 - name: remove 'needs-rebase' label when conflict is resolved
   conditions:
-    - -conflict
-    - -closed
+      - -conflict
+      - -closed
   actions:
     label:
       remove:
         - needs-rebase
-
-- name: label-kv-connector
-  description: Automatically apply kv-connector label
-  conditions:
-    - label != stale
-    - or:
-      - files~=^examples/online_serving/disaggregated[^/]*/.*
-      - files~=^examples/offline_inference/disaggregated[^/]*/.*
-      - files~=^examples/others/lmcache/
-      - files~=^tests/v1/kv_connector/
-      - files~=^vllm/distributed/kv_transfer/
-      - title~=(?i)\bP/?D\b
-      - title~=(?i)NIXL
-      - title~=(?i)LMCache
-  actions:
-    label:
-      add:
-        - kv-connector

.github/workflows/add_label_automerge.yml

@@ -10,7 +10,7 @@ jobs:
         runs-on: ubuntu-latest
         steps:
             -   name: Add label
-                uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+                uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
                 with:
                     script: |
                         github.rest.issues.addLabels({

.github/workflows/bc-lint.yml

@@ -1,29 +0,0 @@
-name: BC Lint
-
-on:
-  pull_request:
-    types:
-      - opened
-      - synchronize
-      - reopened
-      - labeled
-      - unlabeled
-
-jobs:
-  bc_lint:
-    if: github.repository_owner == 'vllm-project'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Run BC Lint Action
-        uses: pytorch/test-infra/.github/actions/bc-lint@main
-        with:
-          repo: ${{ github.event.pull_request.head.repo.full_name }}
-          base_sha: ${{ github.event.pull_request.base.sha }}
-          head_sha: ${{ github.event.pull_request.head.sha }}
-          suppression: ${{ contains(github.event.pull_request.labels.*.name, 'suppress-bc-linter') }}
-          docs_link: 'https://github.com/pytorch/test-infra/wiki/BC-Linter'
-          config_dir: .github
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
-  cancel-in-progress: true

.github/workflows/issue_autolabel.yml

@@ -13,8 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Label issues based on keywords
-        id: label-step
-        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd  # v8.0.0
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea  # v7.0.1
         with:
           script: |
             // Configuration: Add new labels and keywords here
@@ -43,6 +42,7 @@ jobs:
                     searchIn: "body"
                   },
                 ],
+                
                 // Substring search - matches anywhere in text (partial matches)
                 substrings: [
                   {
@@ -89,12 +89,14 @@ jobs:
                     term: "hip_",
                     searchIn: "both"
                   },
+                  
                   // ROCm tools and libraries
                   {
                     term: "hipify",
                     searchIn: "both"
                   },
                 ],
+                
                 // Regex patterns - for complex pattern matching
                 regexPatterns: [
                   {
@@ -105,17 +107,13 @@ jobs:
                   }
                 ],
               },
-              // Add more label configurations here as needed
-              // example: {
-              //   keywords: [...],
-              //   substrings: [...],
-              //   regexPatterns: [...]
-              // },
             };
+            
             // Helper function to create regex based on search type
             function createSearchRegex(term, type) {
               // Escape special regex characters in the term
               const escapedTerm = term.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+              
               switch (type) {
                 case 'keyword':
                   // Word boundary search - matches whole words only
@@ -127,13 +125,16 @@ jobs:
                   throw new Error(`Unknown search type: ${type}`);
               }
             }
+            
             // Helper function to find matching terms in text with line information
             function findMatchingTermsWithLines(text, searchTerms = [], searchType = 'keyword', searchLocation = '') {
               const matches = [];
               const lines = text.split('\n');
+              
               for (const termConfig of searchTerms) {
                 let regex;
                 let term, searchIn, pattern, description, flags;
+                
                 // Handle different input formats (string or object)
                 if (typeof termConfig === 'string') {
                   term = termConfig;
@@ -145,17 +146,21 @@ jobs:
                   description = termConfig.description;
                   flags = termConfig.flags;
                 }
+                
                 // Skip if this term shouldn't be searched in the current location
                 if (searchIn !== 'both' && searchIn !== searchLocation) {
                   continue;
                 }
+                
                 // Create appropriate regex
                 if (searchType === 'regex') {
                   regex = new RegExp(pattern, flags || "gi");
                 } else {
                   regex = createSearchRegex(term, searchType);
                 }
+                
                 const termMatches = [];
+                
                 // Check each line for matches
                 lines.forEach((line, lineIndex) => {
                   const lineMatches = line.match(regex);
@@ -170,14 +175,15 @@ jobs:
                         originalTerm: term || pattern,
                         description: description,
                         // Show context around the match in the line
-                        context: line.length > 100 ?
-                          line.substring(Math.max(0, line.toLowerCase().indexOf(match.toLowerCase()) - 30),
-                                       line.toLowerCase().indexOf(match.toLowerCase()) + match.length + 30) + '...'
+                        context: line.length > 100 ? 
+                          line.substring(Math.max(0, line.toLowerCase().indexOf(match.toLowerCase()) - 30), 
+                                       line.toLowerCase().indexOf(match.toLowerCase()) + match.length + 30) + '...' 
                           : line.trim()
                       });
                     });
                   }
                 });
+                
                 if (termMatches.length > 0) {
                   matches.push({
                     term: term || (description || pattern),
@@ -190,48 +196,64 @@ jobs:
                   });
                 }
               }
+              
               return matches;
             }
+            
             // Helper function to check if label should be added
             async function processLabel(labelName, config) {
               const body = context.payload.issue.body || "";
               const title = context.payload.issue.title || "";
+              
               core.notice(`Processing label: ${labelName}`);
               core.notice(`Issue Title: "${title}"`);
               core.notice(`Issue Body length: ${body.length} characters`);
+              
               let shouldAddLabel = false;
               let allMatches = [];
               let reason = '';
+              
               const keywords = config.keywords || [];
               const substrings = config.substrings || [];
               const regexPatterns = config.regexPatterns || [];
+              
               core.notice(`Searching with ${keywords.length} keywords, ${substrings.length} substrings, and ${regexPatterns.length} regex patterns`);
+              
               // Search in title
               if (title.trim()) {
                 core.notice(`Searching in title: "${title}"`);
+                
                 const titleKeywordMatches = findMatchingTermsWithLines(title, keywords, 'keyword', 'title');
                 const titleSubstringMatches = findMatchingTermsWithLines(title, substrings, 'substring', 'title');
                 const titleRegexMatches = findMatchingTermsWithLines(title, regexPatterns, 'regex', 'title');
+                
                 allMatches.push(...titleKeywordMatches, ...titleSubstringMatches, ...titleRegexMatches);
               }
+              
               // Search in body
               if (body.trim()) {
                 core.notice(`Searching in body (${body.length} characters)`);
+                
                 const bodyKeywordMatches = findMatchingTermsWithLines(body, keywords, 'keyword', 'body');
                 const bodySubstringMatches = findMatchingTermsWithLines(body, substrings, 'substring', 'body');
                 const bodyRegexMatches = findMatchingTermsWithLines(body, regexPatterns, 'regex', 'body');
+                
                 allMatches.push(...bodyKeywordMatches, ...bodySubstringMatches, ...bodyRegexMatches);
               }
+              
               if (allMatches.length > 0) {
                 core.notice(`Found ${allMatches.length} matching term(s):`);
+                
                 for (const termMatch of allMatches) {
                   const locationText = termMatch.searchLocation === 'title' ? 'title' : 'body';
                   const searchInText = termMatch.searchIn === 'both' ? 'both' : termMatch.searchIn;
+                  
                   if (termMatch.searchType === 'regex') {
                     core.notice(`  📍 Regex: "${termMatch.term}" (pattern: ${termMatch.pattern}) found ${termMatch.count} time(s) in ${locationText} (configured to search in: ${searchInText}):`);
                   } else {
                     core.notice(`  📍 Term: "${termMatch.term}" (${termMatch.searchType} search) found ${termMatch.count} time(s) in ${locationText} (configured to search in: ${searchInText}):`);
                   }
+                  
                   // Show details for each match
                   termMatch.matches.forEach((match, index) => {
                     core.notice(`    ${index + 1}. Line ${match.lineNumber} in ${match.searchLocation}: "${match.match}" [${match.searchType}]`);
@@ -244,6 +266,7 @@ jobs:
                     }
                   });
                 }
+                
                 shouldAddLabel = true;
                 const totalMatches = allMatches.reduce((sum, t) => sum + t.count, 0);
                 const titleMatches = allMatches.filter(t => t.searchLocation === 'title').reduce((sum, t) => sum + t.count, 0);
@@ -251,10 +274,13 @@ jobs:
                 const keywordMatches = allMatches.filter(t => t.searchType === 'keyword').reduce((sum, t) => sum + t.count, 0);
                 const substringMatches = allMatches.filter(t => t.searchType === 'substring').reduce((sum, t) => sum + t.count, 0);
                 const regexMatches = allMatches.filter(t => t.searchType === 'regex').reduce((sum, t) => sum + t.count, 0);
+                
                 reason = `Found ${totalMatches} total matches (${titleMatches} in title, ${bodyMatches} in body) - ${keywordMatches} keyword matches, ${substringMatches} substring matches, ${regexMatches} regex matches`;
               }
+              
               core.notice(`Final decision: ${shouldAddLabel ? 'ADD LABEL' : 'DO NOT ADD LABEL'}`);
               core.notice(`Reason: ${reason || 'No matching terms found'}`);
+              
               if (shouldAddLabel) {
                 const existingLabels = context.payload.issue.labels.map(l => l.name);
                 if (!existingLabels.includes(labelName)) {
@@ -270,92 +296,14 @@ jobs:
                 core.notice(`Label "${labelName}" already present.`);
                 return false;
               }
+              
               core.notice(`No matching terms found for label "${labelName}".`);
               return false;
             }
+            
             // Process all configured labels
-            const labelsAddedResults = await Promise.all(
-              Object.entries(labelConfig).map(([labelName, config]) => 
-                processLabel(labelName, config).then(added => ({ labelName, added }))
-              )
-            );
-            
-            const numLabelsAdded = labelsAddedResults.filter(r => r.added).length;
-            core.notice(`Processing complete. ${numLabelsAdded} label(s) added.`);
-            
-            // Return which labels were added for the next step
-            const addedLabels = labelsAddedResults.filter(r => r.added).map(r => r.labelName);
-            core.setOutput('labels_added', JSON.stringify(addedLabels));
-            return addedLabels;
-
-      - name: CC users for labeled issues
-        if: steps.label-step.outputs.labels_added != '[]'
-        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd  # v8.0.0
-        with:
-          script: |
-            // Configuration: Map labels to GitHub users to CC
-            // You can add multiple users per label, and multiple label configurations
-            const ccConfig = {
-              rocm: {
-                users: ['hongxiayang', 'tjtanaa', 'vllmellm'],  // Add more users as needed: ['user1', 'user2', 'user3']
-                message: 'CC {users} for ROCm-related issue'  // {users} will be replaced with @mentions
-              },
-              // Add more label -> user mappings here
-              // Example:
-              // cuda: {
-              //   users: ['user1', 'user2'],
-              //   message: 'CC {users} for CUDA-related issue'
-              // },
-              // performance: {
-              //   users: ['perfexpert'],
-              //   message: 'CC {users} for performance issue'
-              // },
-            };
-            
-            const labelsAdded = JSON.parse('${{ steps.label-step.outputs.labels_added }}');
-            core.notice(`Labels added: ${labelsAdded.join(', ')}`);
-            
-            // Get existing comments to check for already mentioned users
-            const comments = await github.rest.issues.listComments({
-              owner: context.repo.owner,
-              repo: context.repo.repo,
-              issue_number: context.issue.number,
-            });
-            
-            const issueBody = context.payload.issue.body || '';
-            const allExistingText = issueBody + '\n' + comments.data.map(c => c.body).join('\n');
-            
-            // Process each label that was added
-            for (const label of labelsAdded) {
-              if (ccConfig[label]) {
-                const config = ccConfig[label];
-                const usersToMention = [];
-                
-                // Check which users haven't been mentioned yet
-                for (const user of config.users) {
-                  const mentionPattern = new RegExp(`@${user}\\b`, 'i');
-                  if (!mentionPattern.test(allExistingText)) {
-                    usersToMention.push(user);
-                  } else {
-                    core.notice(`@${user} already mentioned for label "${label}", skipping`);
-                  }
-                }
-                
-                // Post comment if there are users to mention
-                if (usersToMention.length > 0) {
-                  const mentions = usersToMention.map(u => `@${u}`).join(' ');
-                  const message = config.message.replace('{users}', mentions);
-                  
-                  await github.rest.issues.createComment({
-                    owner: context.repo.owner,
-                    repo: context.repo.repo,
-                    issue_number: context.issue.number,
-                    body: message
-                  });
-                  
-                  core.notice(`CC comment added for label "${label}": ${mentions}`);
-                } else {
-                  core.notice(`All users for label "${label}" already mentioned, skipping comment`);
-                }
-              }
-            }
+            const processLabels = Object.entries(labelConfig)
+              .map(([labelName, config]) => processLabel(labelName, config));
+            const labelsAdded = await Promise.all(processLabels);
+            const numLabelsAdded = labelsAdded.reduce((x, y) => x + y, 0);
+            core.notice(`Processing complete. ${numLabelsAdded} label(s) added.`);

.github/workflows/reminder_comment.yml

@@ -9,7 +9,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Remind to run full CI on PR
-        uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8.0.0
+        uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
         with:
           script: |
             try {

.github/workflows/stale.yml

@@ -13,7 +13,7 @@ jobs:
       actions: write
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/stale@5f858e3efba33a5ca4407a664cc011ad407f2008 # v10.1.0
+      - uses: actions/stale@5bef64f19d7facfb25b37b414482c7164d639639 # v9.1.0
         with:
           # Increasing this value ensures that changes to this workflow
           # propagate to all issues and PRs in days rather than months

.gitignore

@@ -4,7 +4,7 @@
 # vllm-flash-attn built from source
 vllm/vllm_flash_attn/*
 
-# triton jit
+# triton jit 
 .triton
 
 # Byte-compiled / optimized / DLL files
@@ -94,9 +94,6 @@ ipython_config.py
 # generated files
 **/generated/**
 
-# uv
-uv.lock
-
 # pyenv
 #   For a library or package, you might want to ignore these files since the code is
 #   intended to run in multiple environments; otherwise, check them in:
@@ -180,14 +177,6 @@ cython_debug/
 # VSCode
 .vscode/
 
-# Claude
-CLAUDE.md
-.claude/
-
-# Codex
-AGENTS.md
-.codex/
-
 # DS Store
 .DS_Store
 
@@ -220,4 +209,4 @@ shellcheck*/
 csrc/moe/marlin_moe_wna16/kernel_*
 
 # Ignore ep_kernels_workspace folder
-ep_kernels_workspace/
+ep_kernels_workspace/

.markdownlint.yaml

@@ -4,6 +4,7 @@ MD013: false
 MD024:
   siblings_only: true
 MD033: false
+MD042: false
 MD045: false
 MD046: false
 MD051: false

.pre-commit-config.yaml

@@ -6,19 +6,30 @@ default_stages:
   - manual # Run in CI
 exclude: 'vllm/third_party/.*'
 repos:
-- repo: https://github.com/astral-sh/ruff-pre-commit
-  rev: v0.14.0
+- repo: https://github.com/google/yapf
+  rev: v0.43.0
   hooks:
-  - id: ruff-check
+  - id: yapf
+    args: [--in-place, --verbose]
+    # Keep the same list from yapfignore here to avoid yapf failing without any inputs
+    exclude: '(.buildkite|benchmarks|build|examples)/.*'
+- repo: https://github.com/astral-sh/ruff-pre-commit
+  rev: v0.11.7
+  hooks:
+  - id: ruff
     args: [--output-format, github, --fix]
   - id: ruff-format
+    files: ^(.buildkite|benchmarks|examples)/.*
 - repo: https://github.com/crate-ci/typos
-  rev: v1.38.1
+  rev: v1.35.5
   hooks:
   - id: typos
-    args: [--force-exclude]
+- repo: https://github.com/PyCQA/isort
+  rev: 6.0.1
+  hooks:
+  - id: isort
 - repo: https://github.com/pre-commit/mirrors-clang-format
-  rev: v21.1.2
+  rev: v20.1.3
   hooks:
   - id: clang-format
     exclude: 'csrc/(moe/topk_softmax_kernels.cu|quantization/gguf/(ggml-common.h|dequantize.cuh|vecdotq.cuh|mmq.cuh|mmvq.cuh))|vllm/third_party/.*'
@@ -35,55 +46,61 @@ repos:
   hooks:
   - id: actionlint
 - repo: https://github.com/astral-sh/uv-pre-commit
-  rev: 0.9.1
+  rev: 0.6.17
   hooks:
     - id: pip-compile
-      args: [requirements/test.in, -o, requirements/test.txt, --index-strategy, unsafe-best-match, --torch-backend, cu129, --python-platform, x86_64-manylinux_2_28]
+      args: [requirements/test.in, -o, requirements/test.txt, --index-strategy, unsafe-best-match, --torch-backend, cu128]
       files: ^requirements/test\.(in|txt)$
 - repo: local
   hooks:
   - id: format-torch-nightly-test
     name: reformat nightly_torch_test.txt to be in sync with test.in
     language: python
-    entry: python tools/pre_commit/generate_nightly_torch_test.py
+    entry: python tools/generate_nightly_torch_test.py
     files: ^requirements/test\.(in|txt)$
   - id: mypy-local
-    name: Run mypy locally for lowest supported Python version
-    entry: python tools/pre_commit/mypy.py 0 "3.10"
+    name: Run mypy for local Python installation
+    entry: tools/mypy.sh 0 "local"
+    language: python
+    types: [python]
+    additional_dependencies: &mypy_deps [mypy==1.11.1, types-cachetools, types-setuptools, types-PyYAML, types-requests, pydantic]
     stages: [pre-commit] # Don't run in CI
-    <<: &mypy_common
-      language: python
-      types_or: [python, pyi]
-      require_serial: true
-      additional_dependencies: [mypy==1.11.1, regex, types-cachetools, types-setuptools, types-PyYAML, types-requests, types-torch, pydantic]
+  - id: mypy-3.9 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
+    name: Run mypy for Python 3.9
+    entry: tools/mypy.sh 1 "3.9"
+    language: python
+    types: [python]
+    additional_dependencies: *mypy_deps
+    stages: [manual] # Only run in CI
   - id: mypy-3.10 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
     name: Run mypy for Python 3.10
-    entry: python tools/pre_commit/mypy.py 1 "3.10"
-    <<: *mypy_common
+    entry: tools/mypy.sh 1 "3.10"
+    language: python
+    types: [python]
+    additional_dependencies: *mypy_deps
     stages: [manual] # Only run in CI
   - id: mypy-3.11 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
     name: Run mypy for Python 3.11
-    entry: python tools/pre_commit/mypy.py 1 "3.11"
-    <<: *mypy_common
+    entry: tools/mypy.sh 1 "3.11"
+    language: python
+    types: [python]
+    additional_dependencies: *mypy_deps
     stages: [manual] # Only run in CI
   - id: mypy-3.12 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
     name: Run mypy for Python 3.12
-    entry: python tools/pre_commit/mypy.py 1 "3.12"
-    <<: *mypy_common
-    stages: [manual] # Only run in CI
-  - id: mypy-3.13 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
-    name: Run mypy for Python 3.13
-    entry: python tools/pre_commit/mypy.py 1 "3.13"
-    <<: *mypy_common
+    entry: tools/mypy.sh 1 "3.12"
+    language: python
+    types: [python]
+    additional_dependencies: *mypy_deps
     stages: [manual] # Only run in CI
   - id: shellcheck
     name: Lint shell scripts
-    entry: tools/pre_commit/shellcheck.sh
+    entry: tools/shellcheck.sh
     language: script
     types: [shell]
   - id: png-lint
     name: Lint PNG exports from excalidraw
-    entry: tools/pre_commit/png-lint.sh
+    entry: tools/png-lint.sh
     language: script
     types: [png]
   - id: signoff-commit
@@ -100,12 +117,12 @@ repos:
     stages: [commit-msg]
   - id: check-spdx-header
     name: Check SPDX headers
-    entry: python tools/pre_commit/check_spdx_header.py
+    entry: python tools/check_spdx_header.py
     language: python
     types: [python]
   - id: check-root-lazy-imports
     name: Check root lazy imports
-    entry: python tools/pre_commit/check_init_lazy_imports.py
+    entry: python tools/check_init_lazy_imports.py
     language: python
     types: [python]
   - id: check-filenames
@@ -119,11 +136,11 @@ repos:
     pass_filenames: false
   - id: update-dockerfile-graph
     name: Update Dockerfile dependency graph
-    entry: tools/pre_commit/update-dockerfile-graph.sh
+    entry: tools/update-dockerfile-graph.sh
     language: script
   - id: enforce-import-regex-instead-of-re
     name: Enforce import regex as re
-    entry: python tools/pre_commit/enforce_regex_import.py
+    entry: python tools/enforce_regex_import.py
     language: python
     types: [python]
     pass_filenames: false
@@ -131,22 +148,25 @@ repos:
   # forbid directly import triton
   - id: forbid-direct-triton-import
     name: "Forbid direct 'import triton'"
-    entry: python tools/pre_commit/check_triton_import.py
+    entry: python tools/check_triton_import.py
     language: python
     types: [python]
     pass_filenames: false
     additional_dependencies: [regex]
   - id: check-pickle-imports
     name: Prevent new pickle/cloudpickle imports
-    entry: python tools/pre_commit/check_pickle_imports.py
+    entry: python tools/check_pickle_imports.py
     language: python
     types: [python]
-    additional_dependencies: [regex]
+    pass_filenames: false
+    additional_dependencies: [pathspec, regex]
   - id: validate-config
     name: Validate configuration has default values and that each field has a docstring
-    entry: python tools/pre_commit/validate_config.py
+    entry: python tools/validate_config.py
     language: python
-    additional_dependencies: [regex]
+    types: [python]
+    pass_filenames: true
+    files: vllm/config.py|tests/test_config.py|vllm/entrypoints/openai/cli_args.py
   # Keep `suggestion` last
   - id: suggestion
     name: Suggestion

.readthedocs.yaml

@@ -13,7 +13,6 @@ build:
 
 mkdocs:
   configuration: mkdocs.yaml
-  fail_on_warning: true
 
 # Optionally declare the Python requirements required to build your docs
 python:

.yapfignore

@@ -1,2 +1 @@
 collect_env.py
-vllm/model_executor/layers/fla/ops/*.py

CMakeLists.txt

@@ -13,10 +13,6 @@ cmake_minimum_required(VERSION 3.26)
 # cmake --install . --component _C
 project(vllm_extensions LANGUAGES CXX)
 
-set(CMAKE_CXX_STANDARD 17)
-set(CMAKE_CXX_STANDARD_REQUIRED ON)
-
-
 # CUDA by default, can be overridden by using -DVLLM_TARGET_DEVICE=... (used by setup.py)
 set(VLLM_TARGET_DEVICE "cuda" CACHE STRING "Target device backend for vLLM")
 message(STATUS "Build type: ${CMAKE_BUILD_TYPE}")
@@ -34,10 +30,10 @@ install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
 # Supported python versions.  These versions will be searched in order, the
 # first match will be selected.  These should be kept in sync with setup.py.
 #
-set(PYTHON_SUPPORTED_VERSIONS "3.10" "3.11" "3.12" "3.13")
+set(PYTHON_SUPPORTED_VERSIONS "3.9" "3.10" "3.11" "3.12" "3.13")
 
 # Supported AMD GPU architectures.
-set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151")
+set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201")
 
 #
 # Supported/expected torch versions for CUDA/ROCm.
@@ -49,8 +45,8 @@ set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1
 # requirements.txt files and should be kept consistent.  The ROCm torch
 # versions are derived from docker/Dockerfile.rocm
 #
-set(TORCH_SUPPORTED_VERSION_CUDA "2.9.0")
-set(TORCH_SUPPORTED_VERSION_ROCM "2.9.0")
+set(TORCH_SUPPORTED_VERSION_CUDA "2.8.0")
+set(TORCH_SUPPORTED_VERSION_ROCM "2.8.0")
 
 #
 # Try to find python package with an executable that exactly matches
@@ -86,9 +82,6 @@ find_package(Torch REQUIRED)
 # Supported NVIDIA architectures.
 # This check must happen after find_package(Torch) because that's when CMAKE_CUDA_COMPILER_VERSION gets defined
 if(DEFINED CMAKE_CUDA_COMPILER_VERSION AND
-   CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 13.0)
-  set(CUDA_SUPPORTED_ARCHS "7.5;8.0;8.6;8.7;8.9;9.0;10.0;11.0;12.0")
-elseif(DEFINED CMAKE_CUDA_COMPILER_VERSION AND
    CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.8)
   set(CUDA_SUPPORTED_ARCHS "7.0;7.2;7.5;8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0")
 else()
@@ -178,25 +171,6 @@ if(NVCC_THREADS AND VLLM_GPU_LANG STREQUAL "CUDA")
   list(APPEND VLLM_GPU_FLAGS "--threads=${NVCC_THREADS}")
 endif()
 
-#
-# Set compression mode for CUDA >=13.x.
-#
-if(VLLM_GPU_LANG STREQUAL "CUDA" AND
-   DEFINED CMAKE_CUDA_COMPILER_VERSION AND
-   CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 13.0)
-  list(APPEND VLLM_GPU_FLAGS "--compress-mode=size")
-endif()
-
-#
-# Set CUDA include flags for CXX compiler.
-#
-if(VLLM_GPU_LANG STREQUAL "CUDA")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${CUDA_TOOLKIT_ROOT_DIR}/include")
-  if(CUDA_VERSION VERSION_GREATER_EQUAL 13.0)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${CUDA_TOOLKIT_ROOT_DIR}/include/cccl")
-  endif()
-endif()
-
 #
 # Use FetchContent for C++ dependencies that are compiled as part of vLLM's build process.
 # setup.py will override FETCHCONTENT_BASE_DIR to play nicely with sccache.
@@ -269,8 +243,8 @@ set(VLLM_EXT_SRC
   "csrc/sampler.cu"
   "csrc/cuda_view.cu"
   "csrc/quantization/gptq/q_gemm.cu"
-  "csrc/quantization/w8a8/int8/scaled_quant.cu"
-  "csrc/quantization/w8a8/fp8/common.cu"
+  "csrc/quantization/compressed_tensors/int8_quant_kernels.cu"
+  "csrc/quantization/fp8/common.cu"
   "csrc/quantization/fused_kernels/fused_layernorm_dynamic_per_token_quant.cu"
   "csrc/quantization/gguf/gguf_kernel.cu"
   "csrc/quantization/activation_kernels.cu"
@@ -282,7 +256,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   SET(CUTLASS_ENABLE_HEADERS_ONLY ON CACHE BOOL "Enable only the header library")
 
   # Set CUTLASS_REVISION. Used for FetchContent. Also fixes some bogus messages when building.
-  set(CUTLASS_REVISION "v4.2.1" CACHE STRING "CUTLASS revision to use")
+  set(CUTLASS_REVISION "v4.0.0" CACHE STRING "CUTLASS revision to use")
 
   # Use the specified CUTLASS source directory for compilation if VLLM_CUTLASS_SRC_DIR is provided
   if (DEFINED ENV{VLLM_CUTLASS_SRC_DIR})
@@ -314,13 +288,14 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   list(APPEND VLLM_EXT_SRC
     "csrc/quantization/awq/gemm_kernels.cu"
     "csrc/permute_cols.cu"
-    "csrc/quantization/w8a8/cutlass/scaled_mm_entry.cu"
+    "csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu"
     "csrc/quantization/fp4/nvfp4_quant_entry.cu"
     "csrc/quantization/fp4/nvfp4_scaled_mm_entry.cu"
+    "csrc/quantization/fp4/nvfp4_blockwise_moe_kernel.cu"
     "csrc/sparse/cutlass/sparse_scaled_mm_entry.cu"
     "csrc/cutlass_extensions/common.cpp"
-    "csrc/quantization/w8a8/fp8/per_token_group_quant.cu"
-    "csrc/quantization/w8a8/int8/per_token_group_quant.cu")
+    "csrc/attention/mla/cutlass_mla_entry.cu"
+    "csrc/quantization/fp8/per_token_group_quant.cu")
 
   set_gencode_flags_for_srcs(
     SRCS "${VLLM_EXT_SRC}"
@@ -424,11 +399,11 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.0 AND SCALED_MM_ARCHS)
     set(SRCS
-       "csrc/quantization/w8a8/cutlass/scaled_mm_c3x_sm90.cu"
-       "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_sm90_fp8.cu"
-       "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_sm90_int8.cu"
-       "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_azp_sm90_int8.cu"
-       "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_blockwise_sm90_fp8.cu")
+       "csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm90.cu"
+       "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm90_fp8.cu"
+       "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm90_int8.cu"
+       "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_azp_sm90_int8.cu"
+       "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_blockwise_sm90_fp8.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${SCALED_MM_ARCHS}")
@@ -452,16 +427,12 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 
   # The cutlass_scaled_mm kernels for Geforce Blackwell SM120 (c3x, i.e. CUTLASS 3.x) require
   # CUDA 12.8 or later
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "12.0a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "12.0;12.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
     set(SRCS
-      "csrc/quantization/w8a8/cutlass/scaled_mm_c3x_sm120.cu"
-      "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_sm120_fp8.cu"
-      "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_blockwise_sm120_fp8.cu"
+      "csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm120.cu"
+      "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm120_fp8.cu"
+      "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_blockwise_sm120_fp8.cu"
     )
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
@@ -486,16 +457,12 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 
   # The cutlass_scaled_mm kernels for Blackwell SM100 (c3x, i.e. CUTLASS 3.x)
   # require CUDA 12.8 or later
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0f;11.0f;12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a;10.3a;12.0a;12.1a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
     set(SRCS
-      "csrc/quantization/w8a8/cutlass/scaled_mm_c3x_sm100.cu"
-      "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_sm100_fp8.cu"
-      "csrc/quantization/w8a8/cutlass/c3x/scaled_mm_blockwise_sm100_fp8.cu"
+      "csrc/quantization/cutlass_w8a8/scaled_mm_c3x_sm100.cu"
+      "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_sm100_fp8.cu"
+      "csrc/quantization/cutlass_w8a8/c3x/scaled_mm_blockwise_sm100_fp8.cu"
     )
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
@@ -526,7 +493,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   # subtract out the archs that are already built for 3x
   list(REMOVE_ITEM SCALED_MM_2X_ARCHS ${SCALED_MM_3X_ARCHS})
   if (SCALED_MM_2X_ARCHS)
-    set(SRCS "csrc/quantization/w8a8/cutlass/scaled_mm_c2x.cu")
+    set(SRCS "csrc/quantization/cutlass_w8a8/scaled_mm_c2x.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${SCALED_MM_2X_ARCHS}")
@@ -570,11 +537,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
 
   # The nvfp4_scaled_mm_sm120 kernels for Geforce Blackwell SM120 require
   # CUDA 12.8 or later
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(FP4_ARCHS "12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(FP4_ARCHS "12.0a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(FP4_ARCHS "12.0;12.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
     set(SRCS
       "csrc/quantization/fp4/nvfp4_quant_kernels.cu"
@@ -593,11 +556,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   endif()
 
   # FP4 Archs and flags
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(FP4_ARCHS "10.0f;11.0f;12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(FP4_ARCHS "10.0a;10.1a;12.0a;12.1a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(FP4_ARCHS "10.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND FP4_ARCHS)
     set(SRCS
       "csrc/quantization/fp4/nvfp4_quant_kernels.cu"
@@ -619,13 +578,10 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   endif()
 
   # CUTLASS MLA Archs and flags
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(MLA_ARCHS "10.0f;11.0f;12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(MLA_ARCHS "10.0a;10.1a;10.3a;12.0a;12.1a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(MLA_ARCHS "10.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND MLA_ARCHS)
     set(SRCS
+      "csrc/attention/mla/cutlass_mla_kernels.cu"
       "csrc/attention/mla/sm100_cutlass_mla_kernel.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
@@ -649,7 +605,7 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   # if it's possible to compile MoE kernels that use its output.
   cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
-    set(SRCS "csrc/quantization/w8a8/cutlass/moe/grouped_mm_c3x_sm90.cu")
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x_sm90.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${SCALED_MM_ARCHS}")
@@ -667,13 +623,9 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     endif()
   endif()
 
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0f;11.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
-    set(SRCS "csrc/quantization/w8a8/cutlass/moe/grouped_mm_c3x_sm100.cu")
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x_sm100.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${SCALED_MM_ARCHS}")
@@ -692,13 +644,9 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   endif()
 
   # moe_data.cu is used by all CUTLASS MoE kernels.
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(CUTLASS_MOE_DATA_ARCHS "9.0a;10.0f;11.0f;12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(CUTLASS_MOE_DATA_ARCHS "9.0a;10.0a;10.1a;10.3a;12.0a;12.1a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(CUTLASS_MOE_DATA_ARCHS "9.0a;10.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND CUTLASS_MOE_DATA_ARCHS)
-    set(SRCS "csrc/quantization/w8a8/cutlass/moe/moe_data.cu")
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${CUTLASS_MOE_DATA_ARCHS}")
@@ -715,13 +663,9 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     endif()
   endif()
 
-  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0f;11.0f;12.0f" "${CUDA_ARCHS}")
-  else()
-    cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a;10.1a;10.3a;12.0a;12.1a" "${CUDA_ARCHS}")
-  endif()
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "10.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.8 AND SCALED_MM_ARCHS)
-    set(SRCS "csrc/quantization/w8a8/cutlass/moe/blockwise_scaled_group_mm_sm100.cu")
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/blockwise_scaled_group_mm_sm100.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${SCALED_MM_ARCHS}")
@@ -835,17 +779,6 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     endif()
   endif()
 
-  # Hadacore kernels
-  cuda_archs_loose_intersection(HADACORE_ARCHS "8.0;8.9;9.0" "${CUDA_ARCHS}")
-  if(HADACORE_ARCHS)
-    set(SRCS "csrc/quantization/hadamard/hadacore/hadamard_transform_cuda.cu")
-    set_gencode_flags_for_srcs(
-      SRCS "${SRCS}"
-      CUDA_ARCHS "${HADACORE_ARCHS}")
-    list(APPEND VLLM_EXT_SRC "${SRCS}")
-    message(STATUS "Building hadacore")
-  endif()
-
 # if CUDA endif
 endif()
 
@@ -883,7 +816,6 @@ target_compile_definitions(_C PRIVATE CUTLASS_ENABLE_DIRECT_CUDA_DRIVER_CALL=1)
 set(VLLM_MOE_EXT_SRC
   "csrc/moe/torch_bindings.cpp"
   "csrc/moe/moe_align_sum_kernels.cu"
-  "csrc/moe/moe_lora_align_sum_kernels.cu"
   "csrc/moe/topk_softmax_kernels.cu")
 
 if(VLLM_GPU_LANG STREQUAL "CUDA")
@@ -1008,7 +940,6 @@ endif()
 # For CUDA we also build and ship some external projects.
 if (VLLM_GPU_LANG STREQUAL "CUDA")
     include(cmake/external_projects/flashmla.cmake)
-    include(cmake/external_projects/qutlass.cmake)
 
     # vllm-flash-attn should be last as it overwrites some CMake functions
     include(cmake/external_projects/vllm_flash_attn.cmake)

README.md

@@ -14,15 +14,10 @@ Easy, fast, and cheap LLM serving for everyone
 | <a href="https://docs.vllm.ai"><b>Documentation</b></a> | <a href="https://blog.vllm.ai/"><b>Blog</b></a> | <a href="https://arxiv.org/abs/2309.06180"><b>Paper</b></a> | <a href="https://x.com/vllm_project"><b>Twitter/X</b></a> | <a href="https://discuss.vllm.ai"><b>User Forum</b></a> | <a href="https://slack.vllm.ai"><b>Developer Slack</b></a> |
 </p>
 
----
-Join us at the [PyTorch Conference, October 22-23](https://events.linuxfoundation.org/pytorch-conference/) and [Ray Summit, November 3-5](https://www.anyscale.com/ray-summit/2025) in San Francisco for our latest updates on vLLM and to meet the vLLM team! Register now for the largest vLLM community events of the year!
-
 ---
 
 *Latest News* 🔥
 
-- [2025/10] We hosted [vLLM Shanghai Meetup](https://mp.weixin.qq.com/s/__xb4OyOsImz-9eAVrdlcg) focused on hands-on vLLM inference optimization! Please find the meetup slides [here](https://drive.google.com/drive/folders/1KqwjsFJLfEsC8wlDugnrR61zsWHt94Q6).
-- [2025/09] We hosted [vLLM Toronto Meetup](https://luma.com/e80e0ymm) focused on tackling inference at scale and speculative decoding with speakers from NVIDIA and Red Hat! Please find the meetup slides [here](https://docs.google.com/presentation/d/1IYJYmJcu9fLpID5N5RbW_vO0XLo0CGOR14IXOjB61V8/edit?usp=sharing).
 - [2025/08] We hosted [vLLM Shenzhen Meetup](https://mp.weixin.qq.com/s/k8ZBO1u2_2odgiKWH_GVTQ) focusing on the ecosystem around vLLM! Please find the meetup slides [here](https://drive.google.com/drive/folders/1Ua2SVKVSu-wp5vou_6ElraDt2bnKhiEA).
 - [2025/08] We hosted [vLLM Singapore Meetup](https://www.sginnovate.com/event/vllm-sg-meet). We shared V1 updates, disaggregated serving and MLLM speedups with speakers from Embedded LLM, AMD, WekaIO, and A*STAR. Please find the meetup slides [here](https://drive.google.com/drive/folders/1ncf3GyqLdqFaB6IeB834E5TZJPLAOiXZ?usp=sharing).
 - [2025/08] We hosted [vLLM Shanghai Meetup](https://mp.weixin.qq.com/s/pDmAXHcN7Iqc8sUKgJgGtg) focusing on building, developing, and integrating with vLLM! Please find the meetup slides [here](https://drive.google.com/drive/folders/1OvLx39wnCGy_WKq8SiVKf7YcxxYI3WCH).
@@ -83,7 +78,7 @@ vLLM is flexible and easy to use with:
 - Tensor, pipeline, data and expert parallelism support for distributed inference
 - Streaming outputs
 - OpenAI-compatible API server
-- Support for NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs, and TPU. Additionally, support for diverse hardware plugins such as Intel Gaudi, IBM Spyre and Huawei Ascend.
+- Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs, TPU, and AWS Neuron
 - Prefix caching support
 - Multi-LoRA support
 
@@ -150,7 +145,6 @@ Compute Resources:
 - Trainy
 - UC Berkeley
 - UC San Diego
-- Volcengine
 
 Slack Sponsor: Anyscale
 

benchmarks/README.md

@@ -1,20 +1,807 @@
-# Benchmarks
+# Benchmarking vLLM
 
-This directory used to contain vLLM's benchmark scripts and utilities for performance testing and evaluation.
+This README guides you through running benchmark tests with the extensive
+datasets supported on vLLM. It’s a living document, updated as new features and datasets
+become available.
 
-## Contents
+## Dataset Overview
 
-- **Serving benchmarks**: Scripts for testing online inference performance (latency, throughput)
-- **Throughput benchmarks**: Scripts for testing offline batch inference performance
-- **Specialized benchmarks**: Tools for testing specific features like structured output, prefix caching, long document QA, request prioritization, and multi-modal inference
-- **Dataset utilities**: Framework for loading and sampling from various benchmark datasets (ShareGPT, HuggingFace datasets, synthetic data, etc.)
+<table style="width:100%; border-collapse: collapse;">
+  <thead>
+    <tr>
+      <th style="width:15%; text-align: left;">Dataset</th>
+      <th style="width:10%; text-align: center;">Online</th>
+      <th style="width:10%; text-align: center;">Offline</th>
+      <th style="width:65%; text-align: left;">Data Path</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td><strong>ShareGPT</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json</code></td>
+    </tr>
+    <tr>
+      <td><strong>ShareGPT4V (Image)</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td>
+        <code>wget https://huggingface.co/datasets/Lin-Chen/ShareGPT4V/blob/main/sharegpt4v_instruct_gpt4-vision_cap100k.json</code>
+        <br>
+        <div>Note that the images need to be downloaded separately. For example, to download COCO's 2017 Train images:</div>
+        <code>wget http://images.cocodataset.org/zips/train2017.zip</code>
+      </td>
+    </tr>
+        <tr>
+      <td><strong>ShareGPT4Video (Video)</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td>
+        <code>git clone https://huggingface.co/datasets/ShareGPT4Video/ShareGPT4Video</code>
+      </td>
+    </tr>
+    <tr>
+      <td><strong>BurstGPT</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>wget https://github.com/HPMLL/BurstGPT/releases/download/v1.1/BurstGPT_without_fails_2.csv</code></td>
+    </tr>
+    <tr>
+      <td><strong>Sonnet (deprecated)</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td>Local file: <code>benchmarks/sonnet.txt</code></td>
+    </tr>
+    <tr>
+      <td><strong>Random</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>synthetic</code></td>
+    </tr>
+    <tr>
+      <td><strong>RandomMultiModal (Image/Video)</strong></td>
+      <td style="text-align: center;">🟡</td>
+      <td style="text-align: center;">🚧</td>
+      <td><code>synthetic</code> </td>
+    </tr>
+    <tr>
+      <td><strong>Prefix Repetition</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>synthetic</code></td>
+    </tr>
+    <tr>
+      <td><strong>HuggingFace-VisionArena</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>lmarena-ai/VisionArena-Chat</code></td>
+    </tr>
+    <tr>
+      <td><strong>HuggingFace-InstructCoder</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>likaixin/InstructCoder</code></td>
+    </tr>
+      <tr>
+      <td><strong>HuggingFace-AIMO</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>AI-MO/aimo-validation-aime</code> , <code>AI-MO/NuminaMath-1.5</code>, <code>AI-MO/NuminaMath-CoT</code></td>
+    </tr>
+    <tr>
+      <td><strong>HuggingFace-Other</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td><code>lmms-lab/LLaVA-OneVision-Data</code>, <code>Aeala/ShareGPT_Vicuna_unfiltered</code></td>
+    </tr>
+    <tr>
+      <td><strong>Custom</strong></td>
+      <td style="text-align: center;">✅</td>
+      <td style="text-align: center;">✅</td>
+      <td>Local file: <code>data.jsonl</code></td>
+    </tr>
+  </tbody>
+</table>
 
-## Usage
+✅: supported
 
-For detailed usage instructions, examples, and dataset information, see the [Benchmark CLI documentation](https://docs.vllm.ai/en/latest/contributing/benchmarks.html#benchmark-cli).
+🟡: Partial support
 
-For full CLI reference see:
+🚧: to be supported
 
-- <https://docs.vllm.ai/en/latest/cli/bench/latency.html>
-- <https://docs.vllm.ai/en/latest/cli/bench/serve.html>
-- <https://docs.vllm.ai/en/latest/cli/bench/throughput.html>
+**Note**: HuggingFace dataset's `dataset-name` should be set to `hf`.
+For local `dataset-path`, please set `hf-name` to its Hugging Face ID like
+
+```bash
+--dataset-path /datasets/VisionArena-Chat/ --hf-name lmarena-ai/VisionArena-Chat
+```
+
+## 🚀 Example - Online Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+First start serving your model
+
+```bash
+vllm serve NousResearch/Hermes-3-Llama-3.1-8B
+```
+
+Then run the benchmarking script
+
+```bash
+# download dataset
+# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
+vllm bench serve \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --endpoint /v1/completions \
+  --dataset-name sharegpt \
+  --dataset-path <your data path>/ShareGPT_V3_unfiltered_cleaned_split.json \
+  --num-prompts 10
+```
+
+If successful, you will see the following output
+
+```text
+============ Serving Benchmark Result ============
+Successful requests:                     10
+Benchmark duration (s):                  5.78
+Total input tokens:                      1369
+Total generated tokens:                  2212
+Request throughput (req/s):              1.73
+Output token throughput (tok/s):         382.89
+Total Token throughput (tok/s):          619.85
+---------------Time to First Token----------------
+Mean TTFT (ms):                          71.54
+Median TTFT (ms):                        73.88
+P99 TTFT (ms):                           79.49
+-----Time per Output Token (excl. 1st token)------
+Mean TPOT (ms):                          7.91
+Median TPOT (ms):                        7.96
+P99 TPOT (ms):                           8.03
+---------------Inter-token Latency----------------
+Mean ITL (ms):                           7.74
+Median ITL (ms):                         7.70
+P99 ITL (ms):                            8.39
+==================================================
+```
+
+### Custom Dataset
+
+If the dataset you want to benchmark is not supported yet in vLLM, even then you can benchmark on it using `CustomDataset`. Your data needs to be in `.jsonl` format and needs to have "prompt" field per entry, e.g., data.jsonl
+
+```json
+{"prompt": "What is the capital of India?"}
+{"prompt": "What is the capital of Iran?"}
+{"prompt": "What is the capital of China?"}
+```
+
+```bash
+# start server
+VLLM_USE_V1=1 vllm serve meta-llama/Llama-3.1-8B-Instruct
+```
+
+```bash
+# run benchmarking script
+vllm bench serve --port 9001 --save-result --save-detailed \
+  --backend vllm \
+  --model meta-llama/Llama-3.1-8B-Instruct \
+  --endpoint /v1/completions \
+  --dataset-name custom \
+  --dataset-path <path-to-your-data-jsonl> \
+  --custom-skip-chat-template \
+  --num-prompts 80 \
+  --max-concurrency 1 \
+  --temperature=0.3 \
+  --top-p=0.75 \
+  --result-dir "./log/"
+```
+
+You can skip applying chat template if your data already has it by using `--custom-skip-chat-template`.
+
+### VisionArena Benchmark for Vision Language Models
+
+```bash
+# need a model with vision capability here
+vllm serve Qwen/Qwen2-VL-7B-Instruct
+```
+
+```bash
+vllm bench serve \
+  --backend openai-chat \
+  --endpoint-type openai-chat \
+  --model Qwen/Qwen2-VL-7B-Instruct \
+  --endpoint /v1/chat/completions \
+  --dataset-name hf \
+  --dataset-path lmarena-ai/VisionArena-Chat \
+  --hf-split train \
+  --num-prompts 1000
+```
+
+### InstructCoder Benchmark with Speculative Decoding
+
+``` bash
+VLLM_USE_V1=1 vllm serve meta-llama/Meta-Llama-3-8B-Instruct \
+    --speculative-config $'{"method": "ngram",
+    "num_speculative_tokens": 5, "prompt_lookup_max": 5,
+    "prompt_lookup_min": 2}'
+```
+
+``` bash
+vllm bench serve \
+    --model meta-llama/Meta-Llama-3-8B-Instruct \
+    --dataset-name hf \
+    --dataset-path likaixin/InstructCoder \
+    --num-prompts 2048
+```
+
+### Other HuggingFaceDataset Examples
+
+```bash
+vllm serve Qwen/Qwen2-VL-7B-Instruct
+```
+
+`lmms-lab/LLaVA-OneVision-Data`:
+
+```bash
+vllm bench serve \
+  --backend openai-chat \
+  --endpoint-type openai-chat \
+  --model Qwen/Qwen2-VL-7B-Instruct \
+  --endpoint /v1/chat/completions \
+  --dataset-name hf \
+  --dataset-path lmms-lab/LLaVA-OneVision-Data \
+  --hf-split train \
+  --hf-subset "chart2text(cauldron)" \
+  --num-prompts 10
+```
+
+`Aeala/ShareGPT_Vicuna_unfiltered`:
+
+```bash
+vllm bench serve \
+  --backend openai-chat \
+  --endpoint-type openai-chat \
+  --model Qwen/Qwen2-VL-7B-Instruct \
+  --endpoint /v1/chat/completions \
+  --dataset-name hf \
+  --dataset-path Aeala/ShareGPT_Vicuna_unfiltered \
+  --hf-split train \
+  --num-prompts 10
+```
+
+`AI-MO/aimo-validation-aime`:
+
+``` bash
+vllm bench serve \
+    --model Qwen/QwQ-32B \
+    --dataset-name hf \
+    --dataset-path AI-MO/aimo-validation-aime \
+    --num-prompts 10 \
+    --seed 42
+```
+
+`philschmid/mt-bench`:
+
+``` bash
+vllm bench serve \
+    --model Qwen/QwQ-32B \
+    --dataset-name hf \
+    --dataset-path philschmid/mt-bench \
+    --num-prompts 80
+```
+
+### Running With Sampling Parameters
+
+When using OpenAI-compatible backends such as `vllm`, optional sampling
+parameters can be specified. Example client command:
+
+```bash
+vllm bench serve \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --endpoint /v1/completions \
+  --dataset-name sharegpt \
+  --dataset-path <your data path>/ShareGPT_V3_unfiltered_cleaned_split.json \
+  --top-k 10 \
+  --top-p 0.9 \
+  --temperature 0.5 \
+  --num-prompts 10
+```
+
+### Running With Ramp-Up Request Rate
+
+The benchmark tool also supports ramping up the request rate over the
+duration of the benchmark run. This can be useful for stress testing the
+server or finding the maximum throughput that it can handle, given some latency budget.
+
+Two ramp-up strategies are supported:
+
+- `linear`: Increases the request rate linearly from a start value to an end value.
+- `exponential`: Increases the request rate exponentially.
+
+The following arguments can be used to control the ramp-up:
+
+- `--ramp-up-strategy`: The ramp-up strategy to use (`linear` or `exponential`).
+- `--ramp-up-start-rps`: The request rate at the beginning of the benchmark.
+- `--ramp-up-end-rps`: The request rate at the end of the benchmark.
+
+</details>
+
+## 📈 Example - Offline Throughput Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+```bash
+vllm bench throughput \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset-name sonnet \
+  --dataset-path vllm/benchmarks/sonnet.txt \
+  --num-prompts 10
+```
+
+If successful, you will see the following output
+
+```text
+Throughput: 7.15 requests/s, 4656.00 total tokens/s, 1072.15 output tokens/s
+Total num prompt tokens:  5014
+Total num output tokens:  1500
+```
+
+### VisionArena Benchmark for Vision Language Models
+
+```bash
+vllm bench throughput \
+  --model Qwen/Qwen2-VL-7B-Instruct \
+  --backend vllm-chat \
+  --dataset-name hf \
+  --dataset-path lmarena-ai/VisionArena-Chat \
+  --num-prompts 1000 \
+  --hf-split train
+```
+
+The `num prompt tokens` now includes image token counts
+
+```text
+Throughput: 2.55 requests/s, 4036.92 total tokens/s, 326.90 output tokens/s
+Total num prompt tokens:  14527
+Total num output tokens:  1280
+```
+
+### InstructCoder Benchmark with Speculative Decoding
+
+``` bash
+VLLM_WORKER_MULTIPROC_METHOD=spawn \
+VLLM_USE_V1=1 \
+vllm bench throughput \
+    --dataset-name=hf \
+    --dataset-path=likaixin/InstructCoder \
+    --model=meta-llama/Meta-Llama-3-8B-Instruct \
+    --input-len=1000 \
+    --output-len=100 \
+    --num-prompts=2048 \
+    --async-engine \
+    --speculative-config $'{"method": "ngram",
+    "num_speculative_tokens": 5, "prompt_lookup_max": 5,
+    "prompt_lookup_min": 2}'
+```
+
+```text
+Throughput: 104.77 requests/s, 23836.22 total tokens/s, 10477.10 output tokens/s
+Total num prompt tokens:  261136
+Total num output tokens:  204800
+```
+
+### Other HuggingFaceDataset Examples
+
+`lmms-lab/LLaVA-OneVision-Data`:
+
+```bash
+vllm bench throughput \
+  --model Qwen/Qwen2-VL-7B-Instruct \
+  --backend vllm-chat \
+  --dataset-name hf \
+  --dataset-path lmms-lab/LLaVA-OneVision-Data \
+  --hf-split train \
+  --hf-subset "chart2text(cauldron)" \
+  --num-prompts 10
+```
+
+`Aeala/ShareGPT_Vicuna_unfiltered`:
+
+```bash
+vllm bench throughput \
+  --model Qwen/Qwen2-VL-7B-Instruct \
+  --backend vllm-chat \
+  --dataset-name hf \
+  --dataset-path Aeala/ShareGPT_Vicuna_unfiltered \
+  --hf-split train \
+  --num-prompts 10
+```
+
+`AI-MO/aimo-validation-aime`:
+
+```bash
+vllm bench throughput \
+  --model Qwen/QwQ-32B \
+  --backend vllm \
+  --dataset-name hf \
+  --dataset-path AI-MO/aimo-validation-aime \
+  --hf-split train \
+  --num-prompts 10
+```
+
+Benchmark with LoRA adapters:
+
+``` bash
+# download dataset
+# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
+vllm bench throughput \
+  --model meta-llama/Llama-2-7b-hf \
+  --backend vllm \
+  --dataset_path <your data path>/ShareGPT_V3_unfiltered_cleaned_split.json \
+  --dataset_name sharegpt \
+  --num-prompts 10 \
+  --max-loras 2 \
+  --max-lora-rank 8 \
+  --enable-lora \
+  --lora-path yard1/llama-2-7b-sql-lora-test
+  ```
+
+</details>
+
+## 🛠️ Example - Structured Output Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of structured output generation (JSON, grammar, regex).
+
+### Server Setup
+
+```bash
+vllm serve NousResearch/Hermes-3-Llama-3.1-8B
+```
+
+### JSON Schema Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset json \
+  --structured-output-ratio 1.0 \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### Grammar-based Generation Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset grammar \
+  --structure-type grammar \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### Regex-based Generation Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset regex \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### Choice-based Generation Benchmark
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset choice \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+### XGrammar Benchmark Dataset
+
+```bash
+python3 benchmarks/benchmark_serving_structured_output.py \
+  --backend vllm \
+  --model NousResearch/Hermes-3-Llama-3.1-8B \
+  --dataset xgrammar_bench \
+  --request-rate 10 \
+  --num-prompts 1000
+```
+
+</details>
+
+## 📚 Example - Long Document QA Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of long document question-answering with prefix caching.
+
+### Basic Long Document QA Test
+
+```bash
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 16 \
+  --document-length 2000 \
+  --output-len 50 \
+  --repeat-count 5
+```
+
+### Different Repeat Modes
+
+```bash
+# Random mode (default) - shuffle prompts randomly
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 8 \
+  --document-length 3000 \
+  --repeat-count 3 \
+  --repeat-mode random
+
+# Tile mode - repeat entire prompt list in sequence
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 8 \
+  --document-length 3000 \
+  --repeat-count 3 \
+  --repeat-mode tile
+
+# Interleave mode - repeat each prompt consecutively
+python3 benchmarks/benchmark_long_document_qa_throughput.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-documents 8 \
+  --document-length 3000 \
+  --repeat-count 3 \
+  --repeat-mode interleave
+```
+
+</details>
+
+## 🗂️ Example - Prefix Caching Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the efficiency of automatic prefix caching.
+
+### Fixed Prompt with Prefix Caching
+
+```bash
+python3 benchmarks/benchmark_prefix_caching.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --enable-prefix-caching \
+  --num-prompts 1 \
+  --repeat-count 100 \
+  --input-length-range 128:256
+```
+
+### ShareGPT Dataset with Prefix Caching
+
+```bash
+# download dataset
+# wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
+
+python3 benchmarks/benchmark_prefix_caching.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --dataset-path /path/ShareGPT_V3_unfiltered_cleaned_split.json \
+  --enable-prefix-caching \
+  --num-prompts 20 \
+  --repeat-count 5 \
+  --input-length-range 128:256
+```
+
+### Prefix Repetition Dataset
+
+```bash
+vllm bench serve \
+  --backend openai \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --dataset-name prefix_repetition \
+  --num-prompts 100 \
+  --prefix-repetition-prefix-len 512 \
+  --prefix-repetition-suffix-len 128 \
+  --prefix-repetition-num-prefixes 5 \
+  --prefix-repetition-output-len 128
+```
+
+</details>
+
+## ⚡ Example - Request Prioritization Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of request prioritization in vLLM.
+
+### Basic Prioritization Test
+
+```bash
+python3 benchmarks/benchmark_prioritization.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --input-len 128 \
+  --output-len 64 \
+  --num-prompts 100 \
+  --scheduling-policy priority
+```
+
+### Multiple Sequences per Prompt
+
+```bash
+python3 benchmarks/benchmark_prioritization.py \
+  --model meta-llama/Llama-2-7b-chat-hf \
+  --input-len 128 \
+  --output-len 64 \
+  --num-prompts 100 \
+  --scheduling-policy priority \
+  --n 2
+```
+
+</details>
+
+## 👁️ Example - Multi-Modal Benchmark
+
+<details>
+<summary>Show more</summary>
+
+<br/>
+
+Benchmark the performance of multi-modal requests in vLLM.
+
+### Images (ShareGPT4V)
+
+Start vLLM:
+
+```bash
+python -m vllm.entrypoints.openai.api_server \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dtype bfloat16 \
+  --limit-mm-per-prompt '{"image": 1}' \
+  --allowed-local-media-path /path/to/sharegpt4v/images
+```
+
+Send requests with images:
+
+```bash
+python benchmarks/benchmark_serving.py \
+  --backend openai-chat \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dataset-name sharegpt \
+  --dataset-path /path/to/ShareGPT4V/sharegpt4v_instruct_gpt4-vision_cap100k.json \
+  --num-prompts 100 \
+  --save-result \
+  --result-dir ~/vllm_benchmark_results \
+  --save-detailed \
+  --endpoint /v1/chat/completion
+```
+
+### Videos (ShareGPT4Video)
+
+Start vLLM:
+
+```bash
+python -m vllm.entrypoints.openai.api_server \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dtype bfloat16 \
+  --limit-mm-per-prompt '{"video": 1}' \
+  --allowed-local-media-path /path/to/sharegpt4video/videos
+```
+
+Send requests with videos:
+
+```bash
+python benchmarks/benchmark_serving.py \
+  --backend openai-chat \
+  --model Qwen/Qwen2.5-VL-7B-Instruct \
+  --dataset-name sharegpt \
+  --dataset-path /path/to/ShareGPT4Video/llava_v1_5_mix665k_with_video_chatgpt72k_share4video28k.json \
+  --num-prompts 100 \
+  --save-result \
+  --result-dir ~/vllm_benchmark_results \
+  --save-detailed \
+  --endpoint /v1/chat/completion
+```
+
+### Synthetic Random Images (random-mm)
+
+Generate synthetic image inputs alongside random text prompts to stress-test vision models without external datasets.
+
+Notes:
+
+- Works only with online benchmark via the OpenAI  backend (`--backend openai-chat`) and endpoint `/v1/chat/completions`.
+- Video sampling is not yet implemented.
+
+Start the server (example):
+
+```bash
+vllm serve Qwen/Qwen2.5-VL-3B-Instruct \
+  --dtype bfloat16 \
+  --max-model-len 16384 \
+  --limit-mm-per-prompt '{"image": 3, "video": 0}' \
+  --mm-processor-kwargs max_pixels=1003520
+```
+
+Benchmark. It is recommended to use the flag `--ignore-eos` to simulate real responses. You can set the size of the output via the arg `random-output-len`.
+
+Ex.1: Fixed number of items and a single image resolution, enforcing generation of approx 40 tokens:
+
+```bash
+vllm bench serve \
+  --backend openai-chat \
+  --model Qwen/Qwen2.5-VL-3B-Instruct \
+  --endpoint /v1/chat/completions \
+  --dataset-name random-mm \
+  --num-prompts 100 \
+  --max-concurrency 10 \
+  --random-prefix-len 25 \
+  --random-input-len 300 \
+  --random-output-len 40 \
+  --random-range-ratio 0.2 \
+  --random-mm-base-items-per-request 2 \
+  --random-mm-limit-mm-per-prompt '{"image": 3, "video": 0}' \
+  --random-mm-bucket-config '{(224, 224, 1): 1.0}' \
+  --request-rate inf \
+  --ignore-eos \
+  --seed 42
+```
+
+The number of items per request can be controlled by passing multiple image buckets:
+
+```bash
+  --random-mm-base-items-per-request 2 \
+  --random-mm-num-mm-items-range-ratio 0.5 \
+  --random-mm-limit-mm-per-prompt '{"image": 4, "video": 0}' \
+  --random-mm-bucket-config '{(256, 256, 1): 0.7, (720, 1280, 1): 0.3}' \
+```
+
+Flags specific to `random-mm`:
+
+- `--random-mm-base-items-per-request`: base number of multimodal items per request.
+- `--random-mm-num-mm-items-range-ratio`: vary item count uniformly in the closed integer range [floor(n·(1−r)), ceil(n·(1+r))]. Set r=0 to keep it fixed; r=1 allows 0 items.
+- `--random-mm-limit-mm-per-prompt`: per-modality hard caps, e.g. '{"image": 3, "video": 0}'.
+- `--random-mm-bucket-config`: dict mapping (H, W, T) → probability. Entries with probability 0 are removed; remaining probabilities are renormalized to sum to 1. Use T=1 for images. Set any T>1 for videos (video sampling not yet supported).
+
+Behavioral notes:
+
+- If the requested base item count cannot be satisfied under the provided per-prompt limits, the tool raises an error rather than silently clamping.
+
+How sampling works:
+
+- Determine per-request item count k by sampling uniformly from the integer range defined by `--random-mm-base-items-per-request` and `--random-mm-num-mm-items-range-ratio`, then clamp k to at most the sum of per-modality limits.
+- For each of the k items, sample a bucket (H, W, T) according to the normalized probabilities in `--random-mm-bucket-config`, while tracking how many items of each modality have been added.
+- If a modality (e.g., image) reaches its limit from `--random-mm-limit-mm-per-prompt`, all buckets of that modality are excluded and the remaining bucket probabilities are renormalized before continuing.
+This should be seen as an edge case, and if this behavior can be avoided by setting `--random-mm-limit-mm-per-prompt` to a large number. Note that this might result in errors due to engine config `--limit-mm-per-prompt`.
+- The resulting request contains synthetic image data in `multi_modal_data` (OpenAI Chat format). When `random-mm` is used with the OpenAI Chat backend, prompts remain text and MM content is attached via `multi_modal_data`.
+
+</details>

benchmarks/auto_tune/README.md

@@ -149,70 +149,3 @@ The script follows a systematic process to find the optimal parameters:
 4. **Track Best Result**: Throughout the process, the script tracks the parameter combination that has yielded the highest valid throughput so far.
 
 5. **Profile Collection**: For the best-performing run, the script saves the vLLM profiler output, which can be used for deep-dive performance analysis with tools like TensorBoard.
-
-## Batched `auto_tune`
-
-The `batch_auto_tune.sh` script allows you to run multiple `auto_tune.sh` experiments sequentially from a single configuration file. It iterates through a list of parameter sets, executes `auto_tune.sh` for each, and records the results back into the input file.
-
-### Prerequisites
-
-- **jq**: This script requires `jq` to parse the JSON configuration file.
-- **gcloud**: If you plan to upload results to Google Cloud Storage, the `gcloud` CLI must be installed and authenticated.
-
-### How to Run
-
-1. **Create a JSON configuration file**: Create a file (e.g., `runs_config.json`) containing an array of JSON objects. Each object defines the parameters for a single `auto_tune.sh` run.
-
-2. **Execute the script**:
-
-    ```bash
-    bash batch_auto_tune.sh <path_to_json_file> [gcs_upload_path]
-    ```
-
-    - `<path_to_json_file>`: **Required.** Path to your JSON configuration file.
-    - `[gcs_upload_path]`: **Optional.** A GCS path (e.g., `gs://my-bucket/benchmark-results`) where the detailed results and profiles for each run will be uploaded. If this is empty, the results will be available on the local filesystem (see the log for `RESULT_FILE=/path/to/results/file.txt`).
-
-### Configuration File
-
-The JSON configuration file should contain an array of objects. Each object's keys correspond to the configuration variables for `auto_tune.sh` (see the [Configuration table above](#configuration)). These keys will be converted to uppercase environment variables for each run.
-
-Here is an example `runs_config.json` with two benchmark configurations:
-
-```json
-[
-  {
-    "base": "/home/user",
-    "model": "meta-llama/Llama-3.1-8B-Instruct",
-    "system": "TPU", # OR GPU
-    "tp": 8,
-    "input_len": 128,
-    "output_len": 2048,
-    "max_model_len": 2300,
-    "num_seqs_list": "128 256",
-    "num_batched_tokens_list": "8192 16384"
-  },
-  {
-    "base": "/home/user",
-    "model": "meta-llama/Llama-3.1-70B-Instruct",
-    "system": "TPU", # OR GPU
-    "tp": 8,
-    "input_len": 4000,
-    "output_len": 16,
-    "max_model_len": 4096,
-    "num_seqs_list": "64 128",
-    "num_batched_tokens_list": "4096 8192",
-    "max_latency_allowed_ms": 500
-  }
-]
-```
-
-### Output
-
-The script modifies the input JSON file in place, adding the results of each run to the corresponding object. The following fields are added:
-
-- `run_id`: A unique identifier for the run, derived from the timestamp.
-- `status`: The outcome of the run (`SUCCESS`, `FAILURE`, or `WARNING_NO_RESULT_FILE`).
-- `results`: The content of the `result.txt` file from the `auto_tune.sh` run.
-- `gcs_results`: The GCS URL where the run's artifacts are stored (if a GCS path was provided).
-
-A summary of successful and failed runs is also printed to the console upon completion.

benchmarks/auto_tune/auto_tune.sh

@@ -74,7 +74,7 @@ start_server() {
     local vllm_log=$4
     local profile_dir=$5
 
-    pkill -if "vllm serve" || true
+    pkill -if vllm
 
     # Define the common arguments as a bash array.
     # Each argument and its value are separate elements.
@@ -96,22 +96,17 @@ start_server() {
     # This correctly passes each element as a separate argument.
     if [[ -n "$profile_dir" ]]; then
         # Start server with profiling enabled
-        VLLM_SERVER_DEV_MODE=1 VLLM_TORCH_PROFILER_DIR=$profile_dir \
+        VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 VLLM_TORCH_PROFILER_DIR=$profile_dir \
             vllm serve "${common_args_array[@]}" > "$vllm_log" 2>&1 &
     else
         # Start server without profiling
-        VLLM_SERVER_DEV_MODE=1 \
+        VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 \
             vllm serve "${common_args_array[@]}" > "$vllm_log" 2>&1 &
     fi
-    local server_pid=$!
 
     # wait for 10 minutes...
     server_started=0
     for i in {1..60}; do
-        # This line checks whether the server is still alive or not,
-        # since that we should always have permission to send signal to the server process.
-        kill -0 $server_pid 2> /dev/null || break
-
         RESPONSE=$(curl -s -X GET "http://0.0.0.0:8004/health" -w "%{http_code}" -o /dev/stdout)
         STATUS_CODE=$(echo "$RESPONSE" | tail -n 1)
         if [[ "$STATUS_CODE" -eq 200 ]]; then
@@ -123,7 +118,7 @@ start_server() {
     done
 
     if (( ! server_started )); then
-        echo "server did not start within 10 minutes or crashed. Please check server log at $vllm_log".
+        echo "server did not start within 10 minutes. Please check server log at $vllm_log".
         return 1
     else
         return 0
@@ -139,7 +134,7 @@ run_benchmark() {
     echo "vllm_log: $vllm_log"
     echo
     rm -f $vllm_log
-    pkill -if "vllm serve" || true
+    pkill -if vllm
 
     echo "starting server..."
     # Call start_server without a profile_dir to avoid profiling overhead
@@ -232,7 +227,7 @@ run_benchmark() {
 
     echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput"
 
-    pkill -if "vllm serve" || true
+    pkill -if vllm
     sleep 10
     echo "===================="
     return 0
@@ -308,6 +303,6 @@ if (( $(echo "$best_throughput > 0" | bc -l) )); then
 else
     echo "No configuration met the latency requirements. Skipping final profiling run."
 fi
-pkill -if "vllm serve" || true
+pkill -if vllm
 echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH"
 echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" >> "$RESULT"

benchmarks/auto_tune/batch_auto_tune.sh

@@ -1,128 +0,0 @@
-#!/bin/bash
-
-INPUT_JSON="$1"
-GCS_PATH="$2" # Optional GCS path for uploading results for each run
-
-SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &>/dev/null && pwd)
-AUTOTUNE_SCRIPT="$SCRIPT_DIR/auto_tune.sh"
-
-if [[ -z "$INPUT_JSON" ]]; then
-  echo "Error: Input JSON file not provided."
-  echo "Usage: $0 <path_to_json_file> [gcs_upload_path]"
-  exit 1
-fi
-
-if [[ ! -f "$INPUT_JSON" ]]; then
-  echo "Error: File not found at '$INPUT_JSON'"
-  exit 1
-fi
-
-if ! command -v jq &> /dev/null; then
-    echo "Error: 'jq' command not found. Please install jq to process the JSON input."
-    exit 1
-fi
-
-if [[ -n "$GCS_PATH" ]] && ! command -v gcloud &> /dev/null; then
-    echo "Error: 'gcloud' command not found, but a GCS_PATH was provided."
-    exit 1
-fi
-
-SUCCESS_COUNT=0
-FAILURE_COUNT=0
-FAILED_RUNS=()
-SCRIPT_START_TIME=$(date +%s)
-
-json_content=$(cat "$INPUT_JSON")
-if ! num_runs=$(echo "$json_content" | jq 'length'); then
-  echo "Error: Invalid JSON in $INPUT_JSON. 'jq' failed to get array length." >&2
-  exit 1
-fi
-
-echo "Found $num_runs benchmark configurations in $INPUT_JSON."
-echo "Starting benchmark runs..."
-echo "--------------------------------------------------"
-
-for i in $(seq 0 $(($num_runs - 1))); do
-  run_object=$(echo "$json_content" | jq ".[$i]")
-
-  RUN_START_TIME=$(date +%s)
-  ENV_VARS_ARRAY=()
-  # Dynamically create env vars from the JSON object's keys
-  for key in $(echo "$run_object" | jq -r 'keys_unsorted[]'); do
-    value=$(echo "$run_object" | jq -r ".$key")
-    var_name=$(echo "$key" | tr '[:lower:]' '[:upper:]' | tr -cd 'A-Z0-9_')
-    ENV_VARS_ARRAY+=("${var_name}=${value}")
-  done
-
-  echo "Executing run #$((i+1))/$num_runs with parameters: ${ENV_VARS_ARRAY[*]}"
-
-  # Execute auto_tune.sh and capture output
-  RUN_OUTPUT_FILE=$(mktemp)
-  if env "${ENV_VARS_ARRAY[@]}" bash "$AUTOTUNE_SCRIPT" > >(tee -a "$RUN_OUTPUT_FILE") 2>&1; then
-    STATUS="SUCCESS"
-    ((SUCCESS_COUNT++))
-  else
-    STATUS="FAILURE"
-    ((FAILURE_COUNT++))
-    FAILED_RUNS+=("Run #$((i+1)): $(echo $run_object | jq -c .)")
-  fi
-
-  RUN_OUTPUT=$(<"$RUN_OUTPUT_FILE")
-  rm "$RUN_OUTPUT_FILE"
-
-  # Parse results and optionally upload them to GCS
-  RUN_ID=""
-  RESULTS=""
-  GCS_RESULTS_URL=""
-  if [[ "$STATUS" == "SUCCESS" ]]; then
-    RESULT_FILE_PATH=$(echo "$RUN_OUTPUT" | grep 'RESULT_FILE=' | tail -n 1 | cut -d'=' -f2 | tr -s '/' || true)
-
-    if [[ -n "$RESULT_FILE_PATH" && -f "$RESULT_FILE_PATH" ]]; then
-      RUN_ID=$(basename "$(dirname "$RESULT_FILE_PATH")")
-      RESULT_DIR=$(dirname "$RESULT_FILE_PATH")
-      RESULTS=$(cat "$RESULT_FILE_PATH")
-
-      if [[ -n "$GCS_PATH" ]]; then
-        GCS_RESULTS_URL="${GCS_PATH}/${RUN_ID}"
-        echo "Uploading results to GCS..."
-        if gcloud storage rsync --recursive "$RESULT_DIR/" "$GCS_RESULTS_URL"; then
-          echo "GCS upload successful."
-        else
-          echo "Warning: GCS upload failed for RUN_ID $RUN_ID."
-        fi
-      fi
-    else
-      echo "Warning: Could not find result file for a successful run."
-      STATUS="WARNING_NO_RESULT_FILE"
-    fi
-  fi
-
-  # Add the results back into the JSON object for this run
-  json_content=$(echo "$json_content" | jq --argjson i "$i" --arg run_id "$RUN_ID" --arg status "$STATUS" --arg results "$RESULTS" --arg gcs_results "$GCS_RESULTS_URL" \
-    '.[$i] += {run_id: $run_id, status: $status, results: $results, gcs_results: $gcs_results}')
-
-  RUN_END_TIME=$(date +%s)
-  echo "Run finished in $((RUN_END_TIME - RUN_START_TIME)) seconds. Status: $STATUS"
-  echo "--------------------------------------------------"
-
-  # Save intermediate progress back to the file
-  echo "$json_content" > "$INPUT_JSON.tmp" && mv "$INPUT_JSON.tmp" "$INPUT_JSON"
-
-done
-
-SCRIPT_END_TIME=$(date +%s)
-echo "All benchmark runs completed in $((SCRIPT_END_TIME - SCRIPT_START_TIME)) seconds."
-echo
-echo "====================== SUMMARY ======================"
-echo "Successful runs: $SUCCESS_COUNT"
-echo "Failed runs:     $FAILURE_COUNT"
-echo "==================================================="
-
-if [[ $FAILURE_COUNT -gt 0 ]]; then
-  echo "Details of failed runs (see JSON file for full parameters):"
-  for failed in "${FAILED_RUNS[@]}"; do
-    echo "  - $failed"
-  done
-fi
-
-echo "Updated results have been saved to '$INPUT_JSON'."

benchmarks/backend_request_func.py

@@ -8,6 +8,7 @@ import sys
 import time
 import traceback
 from dataclasses import dataclass, field
+from typing import Optional, Union
 
 import aiohttp
 import huggingface_hub.constants
@@ -27,13 +28,13 @@ class RequestFuncInput:
     prompt_len: int
     output_len: int
     model: str
-    model_name: str | None = None
-    logprobs: int | None = None
-    extra_body: dict | None = None
-    multi_modal_content: dict | list[dict] | None = None
+    model_name: Optional[str] = None
+    logprobs: Optional[int] = None
+    extra_body: Optional[dict] = None
+    multi_modal_content: Optional[dict | list[dict]] = None
     ignore_eos: bool = False
-    language: str | None = None
-    request_id: str | None = None
+    language: Optional[str] = None
+    request_id: Optional[str] = None
 
 
 @dataclass
@@ -51,7 +52,7 @@ class RequestFuncOutput:
 
 async def async_request_tgi(
     request_func_input: RequestFuncInput,
-    pbar: tqdm | None = None,
+    pbar: Optional[tqdm] = None,
 ) -> RequestFuncOutput:
     api_url = request_func_input.api_url
     assert api_url.endswith("generate_stream")
@@ -132,7 +133,7 @@ async def async_request_tgi(
 
 async def async_request_trt_llm(
     request_func_input: RequestFuncInput,
-    pbar: tqdm | None = None,
+    pbar: Optional[tqdm] = None,
 ) -> RequestFuncOutput:
     api_url = request_func_input.api_url
     assert api_url.endswith("generate_stream")
@@ -203,7 +204,7 @@ async def async_request_trt_llm(
 
 async def async_request_deepspeed_mii(
     request_func_input: RequestFuncInput,
-    pbar: tqdm | None = None,
+    pbar: Optional[tqdm] = None,
 ) -> RequestFuncOutput:
     api_url = request_func_input.api_url
     assert api_url.endswith(("completions", "profile")), (
@@ -266,7 +267,7 @@ async def async_request_deepspeed_mii(
 
 async def async_request_openai_completions(
     request_func_input: RequestFuncInput,
-    pbar: tqdm | None = None,
+    pbar: Optional[tqdm] = None,
 ) -> RequestFuncOutput:
     api_url = request_func_input.api_url
     assert api_url.endswith(("completions", "profile")), (
@@ -366,7 +367,7 @@ async def async_request_openai_completions(
 
 async def async_request_openai_chat_completions(
     request_func_input: RequestFuncInput,
-    pbar: tqdm | None = None,
+    pbar: Optional[tqdm] = None,
 ) -> RequestFuncOutput:
     api_url = request_func_input.api_url
     assert api_url.endswith(("chat/completions", "profile")), (
@@ -475,7 +476,7 @@ async def async_request_openai_chat_completions(
 
 async def async_request_openai_audio(
     request_func_input: RequestFuncInput,
-    pbar: tqdm | None = None,
+    pbar: Optional[tqdm] = None,
 ) -> RequestFuncOutput:
     # Lazy import without PlaceholderModule to avoid vllm dep.
     import soundfile
@@ -609,7 +610,7 @@ def get_tokenizer(
     tokenizer_mode: str = "auto",
     trust_remote_code: bool = False,
     **kwargs,
-) -> PreTrainedTokenizer | PreTrainedTokenizerFast:
+) -> Union[PreTrainedTokenizer, PreTrainedTokenizerFast]:
     if pretrained_model_name_or_path is not None and not os.path.exists(
         pretrained_model_name_or_path
     ):

benchmarks/benchmark_block_pool.py

@@ -2,10 +2,10 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 import gc
 
-from benchmark_utils import TimeCollector
 from tabulate import tabulate
 
-from vllm.utils.argparse_utils import FlexibleArgumentParser
+from benchmark_utils import TimeCollector
+from vllm.utils import FlexibleArgumentParser
 from vllm.v1.core.block_pool import BlockPool
 
 

benchmarks/benchmark_latency.py

@@ -1,17 +1,191 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-import sys
+"""Benchmark the latency of processing a single batch of requests."""
+
+import argparse
+import dataclasses
+import json
+import os
+import time
+from typing import Any, Optional
+
+import numpy as np
+from tqdm import tqdm
+from typing_extensions import deprecated
+
+import vllm.envs as envs
+from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
+from vllm import LLM, SamplingParams
+from vllm.engine.arg_utils import EngineArgs
+from vllm.inputs import PromptType
+from vllm.sampling_params import BeamSearchParams
+from vllm.utils import FlexibleArgumentParser
+
+
+def save_to_pytorch_benchmark_format(
+    args: argparse.Namespace, results: dict[str, Any]
+) -> None:
+    pt_records = convert_to_pytorch_benchmark_format(
+        args=args,
+        metrics={"latency": results["latencies"]},
+        extra_info={k: results[k] for k in ["avg_latency", "percentiles"]},
+    )
+    if pt_records:
+        pt_file = f"{os.path.splitext(args.output_json)[0]}.pytorch.json"
+        write_to_json(pt_file, pt_records)
+
+
+@deprecated(
+    "benchmark_latency.py is deprecated and will be removed in a "
+    "future version. Please use 'vllm bench latency' instead.",
+)
+def main(args: argparse.Namespace):
+    print(args)
+
+    engine_args = EngineArgs.from_cli_args(args)
+
+    # NOTE(woosuk): If the request cannot be processed in a single batch,
+    # the engine will automatically process the request in multiple batches.
+    llm = LLM(**dataclasses.asdict(engine_args))
+    assert llm.llm_engine.model_config.max_model_len >= (
+        args.input_len + args.output_len
+    ), (
+        "Please ensure that max_model_len is greater than"
+        " the sum of input_len and output_len."
+    )
+
+    sampling_params = SamplingParams(
+        n=args.n,
+        temperature=1.0,
+        top_p=1.0,
+        ignore_eos=True,
+        max_tokens=args.output_len,
+        detokenize=not args.disable_detokenize,
+    )
+    print(sampling_params)
+    dummy_prompt_token_ids = np.random.randint(
+        10000, size=(args.batch_size, args.input_len)
+    )
+    dummy_prompts: list[PromptType] = [
+        {"prompt_token_ids": batch} for batch in dummy_prompt_token_ids.tolist()
+    ]
+
+    def llm_generate():
+        if not args.use_beam_search:
+            llm.generate(dummy_prompts, sampling_params=sampling_params, use_tqdm=False)
+        else:
+            llm.beam_search(
+                dummy_prompts,
+                BeamSearchParams(
+                    beam_width=args.n,
+                    max_tokens=args.output_len,
+                    ignore_eos=True,
+                ),
+            )
+
+    def run_to_completion(profile_dir: Optional[str] = None):
+        if profile_dir:
+            llm.start_profile()
+            llm_generate()
+            llm.stop_profile()
+        else:
+            start_time = time.perf_counter()
+            llm_generate()
+            end_time = time.perf_counter()
+            latency = end_time - start_time
+            return latency
+
+    print("Warming up...")
+    for _ in tqdm(range(args.num_iters_warmup), desc="Warmup iterations"):
+        run_to_completion(profile_dir=None)
+
+    if args.profile:
+        profile_dir = envs.VLLM_TORCH_PROFILER_DIR
+        print(f"Profiling (results will be saved to '{profile_dir}')...")
+        run_to_completion(profile_dir=profile_dir)
+        return
+
+    # Benchmark.
+    latencies = []
+    for _ in tqdm(range(args.num_iters), desc="Profiling iterations"):
+        latencies.append(run_to_completion(profile_dir=None))
+    latencies = np.array(latencies)
+    percentages = [10, 25, 50, 75, 90, 99]
+    percentiles = np.percentile(latencies, percentages)
+    print(f"Avg latency: {np.mean(latencies)} seconds")
+    for percentage, percentile in zip(percentages, percentiles):
+        print(f"{percentage}% percentile latency: {percentile} seconds")
+
+    # Output JSON results if specified
+    if args.output_json:
+        results = {
+            "avg_latency": np.mean(latencies),
+            "latencies": latencies.tolist(),
+            "percentiles": dict(zip(percentages, percentiles.tolist())),
+        }
+        with open(args.output_json, "w") as f:
+            json.dump(results, f, indent=4)
+        save_to_pytorch_benchmark_format(args, results)
+
+
+def create_argument_parser():
+    parser = FlexibleArgumentParser(
+        description="Benchmark the latency of processing a single batch of "
+        "requests till completion."
+    )
+    parser.add_argument("--input-len", type=int, default=32)
+    parser.add_argument("--output-len", type=int, default=128)
+    parser.add_argument("--batch-size", type=int, default=8)
+    parser.add_argument(
+        "--n",
+        type=int,
+        default=1,
+        help="Number of generated sequences per prompt.",
+    )
+    parser.add_argument("--use-beam-search", action="store_true")
+    parser.add_argument(
+        "--num-iters-warmup",
+        type=int,
+        default=10,
+        help="Number of iterations to run for warmup.",
+    )
+    parser.add_argument(
+        "--num-iters", type=int, default=30, help="Number of iterations to run."
+    )
+    parser.add_argument(
+        "--profile",
+        action="store_true",
+        help="profile the generation process of a single batch",
+    )
+    parser.add_argument(
+        "--output-json",
+        type=str,
+        default=None,
+        help="Path to save the latency results in JSON format.",
+    )
+    parser.add_argument(
+        "--disable-detokenize",
+        action="store_true",
+        help=(
+            "Do not detokenize responses (i.e. do not include "
+            "detokenization time in the latency measurement)"
+        ),
+    )
+
+    parser = EngineArgs.add_cli_args(parser)
+    # V1 enables prefix caching by default which skews the latency
+    # numbers. We need to disable prefix caching by default.
+    parser.set_defaults(enable_prefix_caching=False)
+
+    return parser
+
 
 if __name__ == "__main__":
-    print("""DEPRECATED: This script has been moved to the vLLM CLI.
-
-Please use the following command instead:
-    vllm bench latency
-
-For help with the new command, run:
-    vllm bench latency --help
-
-Alternatively, you can run the new command directly with:
-    python -m vllm.entrypoints.cli.main bench latency --help
-""")
-    sys.exit(1)
+    parser = create_argument_parser()
+    args = parser.parse_args()
+    if args.profile and not envs.VLLM_TORCH_PROFILER_DIR:
+        raise OSError(
+            "The environment variable 'VLLM_TORCH_PROFILER_DIR' is not set. "
+            "Please set it to a valid path to use torch profiler."
+        )
+    main(args)

benchmarks/benchmark_long_document_qa_throughput.py

@@ -46,7 +46,7 @@ import time
 
 from vllm import LLM, SamplingParams
 from vllm.engine.arg_utils import EngineArgs
-from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils import FlexibleArgumentParser
 
 
 def test_long_document_qa(llm=None, sampling_params=None, prompts=None):

benchmarks/benchmark_ngram_proposer.py

@@ -1,31 +1,17 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 import gc
-import time
-from unittest import mock
 
 import numpy as np
-from benchmark_utils import TimeCollector
 from tabulate import tabulate
 
-from vllm.config import (
-    CacheConfig,
-    DeviceConfig,
-    LoadConfig,
-    ModelConfig,
-    ParallelConfig,
-    SchedulerConfig,
-    SpeculativeConfig,
-    VllmConfig,
-)
-from vllm.platforms import current_platform
-from vllm.utils.argparse_utils import FlexibleArgumentParser
+from benchmark_utils import TimeCollector
+from vllm.config import ModelConfig, SpeculativeConfig, VllmConfig
+from vllm.utils import FlexibleArgumentParser
 from vllm.v1.spec_decode.ngram_proposer import NgramProposer
-from vllm.v1.worker.gpu_input_batch import InputBatch
-from vllm.v1.worker.gpu_model_runner import GPUModelRunner
 
 
-def benchmark_propose(args):
+def main(args):
     rows = []
     for max_ngram in args.max_ngram:
         collector = TimeCollector(TimeCollector.US)
@@ -83,88 +69,10 @@ def benchmark_propose(args):
     )
 
 
-def benchmark_batched_propose(args):
-    NUM_SPECULATIVE_TOKENS_NGRAM = 10
-    PROMPT_LOOKUP_MIN = 5
-    PROMPT_LOOKUP_MAX = 15
-    MAX_MODEL_LEN = int(1e7)
-    DEVICE = current_platform.device_type
-
-    model_config = ModelConfig(model="facebook/opt-125m", runner="generate")
-
-    speculative_config = SpeculativeConfig(
-        target_model_config=model_config,
-        target_parallel_config=ParallelConfig(),
-        method="ngram",
-        num_speculative_tokens=NUM_SPECULATIVE_TOKENS_NGRAM,
-        prompt_lookup_max=PROMPT_LOOKUP_MAX,
-        prompt_lookup_min=PROMPT_LOOKUP_MIN,
-    )
-
-    vllm_config = VllmConfig(
-        model_config=model_config,
-        cache_config=CacheConfig(),
-        speculative_config=speculative_config,
-        device_config=DeviceConfig(device=current_platform.device_type),
-        parallel_config=ParallelConfig(),
-        load_config=LoadConfig(),
-        scheduler_config=SchedulerConfig(),
-    )
-
-    # monkey patch vllm.v1.worker.gpu_model_runner.get_pp_group
-    mock_pp_group = mock.MagicMock()
-    mock_pp_group.world_size = 1
-    with mock.patch(
-        "vllm.v1.worker.gpu_model_runner.get_pp_group", return_value=mock_pp_group
-    ):
-        runner = GPUModelRunner(vllm_config, DEVICE)
-
-        # hack max model len
-        runner.max_model_len = MAX_MODEL_LEN
-        runner.drafter.max_model_len = MAX_MODEL_LEN
-
-        dummy_input_batch = InputBatch(
-            max_num_reqs=args.num_req,
-            max_model_len=MAX_MODEL_LEN,
-            max_num_batched_tokens=args.num_req * args.num_token,
-            device=DEVICE,
-            pin_memory=False,
-            vocab_size=256000,
-            block_sizes=[16],
-        )
-        dummy_input_batch._req_ids = list(str(id) for id in range(args.num_req))
-        dummy_input_batch.spec_decode_unsupported_reqs = ()
-        dummy_input_batch.num_tokens_no_spec = [args.num_token] * args.num_req
-        dummy_input_batch.token_ids_cpu = np.random.randint(
-            0, 20, (args.num_req, args.num_token)
-        )
-
-        runner.input_batch = dummy_input_batch
-
-        sampled_token_ids = [[0]] * args.num_req
-
-        print("Starting benchmark")
-        # first run is warmup so ignore it
-        for _ in range(args.num_iteration):
-            start = time.time()
-            runner.drafter.propose(
-                sampled_token_ids,
-                dummy_input_batch.req_ids,
-                dummy_input_batch.num_tokens_no_spec,
-                dummy_input_batch.token_ids_cpu,
-                dummy_input_batch.spec_decode_unsupported_reqs,
-            )
-            end = time.time()
-            print(f"Iteration time (s): {end - start}")
-
-
 def invoke_main() -> None:
     parser = FlexibleArgumentParser(
         description="Benchmark the performance of N-gram speculative decode drafting"
     )
-    parser.add_argument(
-        "--batched", action="store_true", help="consider time to prepare batch"
-    )
     parser.add_argument(
         "--num-iteration",
         type=int,
@@ -197,17 +105,8 @@ def invoke_main() -> None:
         help="Number of speculative tokens to generate",
     )
     args = parser.parse_args()
-
-    if not args.batched:
-        benchmark_propose(args)
-    else:
-        benchmark_batched_propose(args)
+    main(args)
 
 
-"""
-# Example command lines:
-# time python3 benchmarks/benchmark_ngram_proposer.py
-# time python3 benchmarks/benchmark_ngram_proposer.py --batched --num-iteration 4 --num-token 1000000 --num-req 128
-"""  # noqa: E501
 if __name__ == "__main__":
     invoke_main()  # pragma: no cover

benchmarks/benchmark_prefix_caching.py

@@ -32,12 +32,13 @@ import dataclasses
 import json
 import random
 import time
+from typing import Optional
 
 from transformers import PreTrainedTokenizerBase
 
 from vllm import LLM, SamplingParams
 from vllm.engine.arg_utils import EngineArgs
-from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils import FlexibleArgumentParser
 
 try:
     from vllm.transformers_utils.tokenizer import get_tokenizer
@@ -79,7 +80,7 @@ def sample_requests_from_dataset(
     num_requests: int,
     tokenizer: PreTrainedTokenizerBase,
     input_length_range: tuple[int, int],
-    fixed_output_len: int | None,
+    fixed_output_len: Optional[int],
 ) -> list[Request]:
     if fixed_output_len is not None and fixed_output_len < 4:
         raise ValueError("output_len too small")
@@ -127,7 +128,7 @@ def sample_requests_from_random(
     num_requests: int,
     tokenizer: PreTrainedTokenizerBase,
     input_length_range: tuple[int, int],
-    fixed_output_len: int | None,
+    fixed_output_len: Optional[int],
     prefix_len: int,
 ) -> list[Request]:
     requests = []

benchmarks/benchmark_prioritization.py

@@ -7,11 +7,12 @@ import dataclasses
 import json
 import random
 import time
+from typing import Optional
 
 from transformers import AutoTokenizer, PreTrainedTokenizerBase
 
 from vllm.engine.arg_utils import EngineArgs
-from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils import FlexibleArgumentParser
 
 
 # Select a equi-probable random priority
@@ -23,7 +24,7 @@ def sample_requests(
     dataset_path: str,
     num_requests: int,
     tokenizer: PreTrainedTokenizerBase,
-    fixed_output_len: int | None,
+    fixed_output_len: Optional[int],
 ) -> list[tuple[str, int, int, int]]:
     if fixed_output_len is not None and fixed_output_len < 4:
         raise ValueError("output_len too small")

benchmarks/benchmark_serving_structured_output.py

@@ -31,19 +31,20 @@ import time
 import uuid
 import warnings
 from collections.abc import AsyncGenerator
-from contextlib import nullcontext
 from dataclasses import dataclass
+from typing import Optional
 
 import datasets
 import numpy as np
 import pandas as pd
+from tqdm.asyncio import tqdm
+from transformers import PreTrainedTokenizerBase
+
 from backend_request_func import (
     ASYNC_REQUEST_FUNCS,
     RequestFuncInput,
     RequestFuncOutput,
 )
-from tqdm.asyncio import tqdm
-from transformers import PreTrainedTokenizerBase
 
 try:
     from vllm.transformers_utils.tokenizer import get_tokenizer
@@ -51,7 +52,7 @@ except ImportError:
     from backend_request_func import get_tokenizer
 
 try:
-    from vllm.utils.argparse_utils import FlexibleArgumentParser
+    from vllm.utils import FlexibleArgumentParser
 except ImportError:
     from argparse import ArgumentParser as FlexibleArgumentParser
 
@@ -316,7 +317,7 @@ def calculate_metrics(
     tokenizer: PreTrainedTokenizerBase,
     selected_percentile_metrics: list[str],
     selected_percentiles: list[float],
-    goodput_config_dict: dict[str, float] | None = None,
+    goodput_config_dict: Optional[dict[str, float]] = None,
 ) -> tuple[BenchmarkMetrics, list[int]]:
     actual_output_lens: list[int] = []
     total_input = 0
@@ -436,9 +437,9 @@ async def benchmark(
     selected_percentile_metrics: list[str],
     selected_percentiles: list[str],
     ignore_eos: bool,
-    max_concurrency: int | None,
+    max_concurrency: Optional[int],
     structured_output_ratio: float,
-    goodput_config_dict: dict[str, float] | None = None,
+    goodput_config_dict: Optional[dict[str, float]] = None,
 ):
     if backend in ASYNC_REQUEST_FUNCS:
         request_func = ASYNC_REQUEST_FUNCS[backend]
@@ -448,8 +449,7 @@ async def benchmark(
     def prepare_extra_body(request) -> dict:
         extra_body = {}
         # Add the schema to the extra_body
-        extra_body["structured_outputs"] = {}
-        extra_body["structured_outputs"][request.structure_type] = request.schema
+        extra_body[request.structure_type] = request.schema
         return extra_body
 
     print("Starting initial single prompt test run...")
@@ -502,9 +502,15 @@ async def benchmark(
 
     pbar = None if disable_tqdm else tqdm(total=len(input_requests))
 
-    semaphore = asyncio.Semaphore(max_concurrency) if max_concurrency else nullcontext()
+    # This can be used once the minimum Python version is 3.10 or higher,
+    # and it will simplify the code in limited_request_func.
+    #    semaphore = (asyncio.Semaphore(max_concurrency)
+    #                 if max_concurrency else contextlib.nullcontext())
+    semaphore = asyncio.Semaphore(max_concurrency) if max_concurrency else None
 
     async def limited_request_func(request_func_input, pbar):
+        if semaphore is None:
+            return await request_func(request_func_input=request_func_input, pbar=pbar)
         async with semaphore:
             return await request_func(request_func_input=request_func_input, pbar=pbar)
 
@@ -690,11 +696,11 @@ def evaluate(ret, args):
         return re.match(args.regex, actual) is not None
 
     def _eval_correctness(expected, actual):
-        if args.structure_type == "json":
+        if args.structure_type == "guided_json":
             return _eval_correctness_json(expected, actual)
-        elif args.structure_type == "regex":
+        elif args.structure_type == "guided_regex":
             return _eval_correctness_regex(expected, actual)
-        elif args.structure_type == "choice":
+        elif args.structure_type == "guided_choice":
             return _eval_correctness_choice(expected, actual)
         else:
             return None
@@ -774,18 +780,18 @@ def main(args: argparse.Namespace):
     )
 
     if args.dataset == "grammar":
-        args.structure_type = "grammar"
+        args.structure_type = "guided_grammar"
     elif args.dataset == "regex":
-        args.structure_type = "regex"
+        args.structure_type = "guided_regex"
     elif args.dataset == "choice":
-        args.structure_type = "choice"
+        args.structure_type = "guided_choice"
     else:
-        args.structure_type = "json"
+        args.structure_type = "guided_json"
 
     if args.no_structured_output:
         args.structured_output_ratio = 0
     if args.save_results:
-        result_file_name = f"{args.structured_output_ratio}so"
+        result_file_name = f"{args.structured_output_ratio}guided"
         result_file_name += f"_{backend}"
         result_file_name += f"_{args.request_rate}qps"
         result_file_name += f"_{args.model.split('/')[-1]}"
@@ -903,13 +909,13 @@ def create_argument_parser():
     parser.add_argument(
         "--tokenizer",
         type=str,
-        help="Name or path of the tokenizer, if not using the default tokenizer.",
+        help="Name or path of the tokenizer, if not using the default tokenizer.",  # noqa: E501
     )
     parser.add_argument(
         "--tokenizer-mode",
         type=str,
         default="auto",
-        help="Name or path of the tokenizer, if not using the default tokenizer.",
+        help="Name or path of the tokenizer, if not using the default tokenizer.",  # noqa: E501
     )
     parser.add_argument(
         "--num-prompts",

benchmarks/benchmark_throughput.py

@@ -1,17 +1,741 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-import sys
+"""Benchmark offline inference throughput."""
+
+import argparse
+import dataclasses
+import json
+import os
+import random
+import time
+import warnings
+from typing import Any, Optional, Union
+
+import torch
+import uvloop
+from tqdm import tqdm
+from transformers import AutoModelForCausalLM, AutoTokenizer, PreTrainedTokenizerBase
+from typing_extensions import deprecated
+
+from benchmark_dataset import (
+    AIMODataset,
+    BurstGPTDataset,
+    ConversationDataset,
+    InstructCoderDataset,
+    RandomDataset,
+    SampleRequest,
+    ShareGPTDataset,
+    SonnetDataset,
+    VisionArenaDataset,
+)
+from benchmark_utils import convert_to_pytorch_benchmark_format, write_to_json
+from vllm.engine.arg_utils import AsyncEngineArgs, EngineArgs
+from vllm.entrypoints.openai.api_server import (
+    build_async_engine_client_from_engine_args,
+)
+from vllm.inputs import TextPrompt, TokensPrompt
+from vllm.lora.request import LoRARequest
+from vllm.outputs import RequestOutput
+from vllm.sampling_params import BeamSearchParams
+from vllm.utils import FlexibleArgumentParser, merge_async_iterators
+
+
+def run_vllm(
+    requests: list[SampleRequest],
+    n: int,
+    engine_args: EngineArgs,
+    disable_detokenize: bool = False,
+) -> tuple[float, Optional[list[RequestOutput]]]:
+    from vllm import LLM, SamplingParams
+
+    llm = LLM(**dataclasses.asdict(engine_args))
+    assert all(
+        llm.llm_engine.model_config.max_model_len
+        >= (request.prompt_len + request.expected_output_len)
+        for request in requests
+    ), (
+        "Please ensure that max_model_len is greater than the sum of"
+        " prompt_len and expected_output_len for all requests."
+    )
+    # Add the requests to the engine.
+    prompts: list[Union[TextPrompt, TokensPrompt]] = []
+    sampling_params: list[SamplingParams] = []
+    for request in requests:
+        prompts.append(
+            TokensPrompt(
+                prompt_token_ids=request.prompt["prompt_token_ids"],
+                multi_modal_data=request.multi_modal_data,
+            )
+            if "prompt_token_ids" in request.prompt
+            else TextPrompt(
+                prompt=request.prompt, multi_modal_data=request.multi_modal_data
+            )
+        )
+        sampling_params.append(
+            SamplingParams(
+                n=n,
+                temperature=1.0,
+                top_p=1.0,
+                ignore_eos=True,
+                max_tokens=request.expected_output_len,
+                detokenize=not disable_detokenize,
+            )
+        )
+    lora_requests: Optional[list[LoRARequest]] = None
+    if engine_args.enable_lora:
+        lora_requests = [request.lora_request for request in requests]
+
+    use_beam_search = False
+
+    outputs = None
+    if not use_beam_search:
+        start = time.perf_counter()
+        outputs = llm.generate(
+            prompts, sampling_params, lora_request=lora_requests, use_tqdm=True
+        )
+        end = time.perf_counter()
+    else:
+        assert lora_requests is None, "BeamSearch API does not support LoRA"
+        # output_len should be the same for all requests.
+        output_len = requests[0].expected_output_len
+        for request in requests:
+            assert request.expected_output_len == output_len
+        start = time.perf_counter()
+        llm.beam_search(
+            prompts,
+            BeamSearchParams(
+                beam_width=n,
+                max_tokens=output_len,
+                ignore_eos=True,
+            ),
+        )
+        end = time.perf_counter()
+    return end - start, outputs
+
+
+def run_vllm_chat(
+    requests: list[SampleRequest],
+    n: int,
+    engine_args: EngineArgs,
+    disable_detokenize: bool = False,
+) -> tuple[float, list[RequestOutput]]:
+    """
+    Run vLLM chat benchmark. This function is recommended ONLY for benchmarking
+    multimodal models as it properly handles multimodal inputs and chat
+    formatting. For non-multimodal models, use run_vllm() instead.
+    """
+    from vllm import LLM, SamplingParams
+
+    llm = LLM(**dataclasses.asdict(engine_args))
+
+    assert all(
+        llm.llm_engine.model_config.max_model_len
+        >= (request.prompt_len + request.expected_output_len)
+        for request in requests
+    ), (
+        "Please ensure that max_model_len is greater than the sum of "
+        "prompt_len and expected_output_len for all requests."
+    )
+
+    prompts = []
+    sampling_params: list[SamplingParams] = []
+    for request in requests:
+        prompts.append(request.prompt)
+        sampling_params.append(
+            SamplingParams(
+                n=n,
+                temperature=1.0,
+                top_p=1.0,
+                ignore_eos=True,
+                max_tokens=request.expected_output_len,
+                detokenize=not disable_detokenize,
+            )
+        )
+    start = time.perf_counter()
+    outputs = llm.chat(prompts, sampling_params, use_tqdm=True)
+    end = time.perf_counter()
+    return end - start, outputs
+
+
+async def run_vllm_async(
+    requests: list[SampleRequest],
+    n: int,
+    engine_args: AsyncEngineArgs,
+    disable_frontend_multiprocessing: bool = False,
+    disable_detokenize: bool = False,
+) -> float:
+    from vllm import SamplingParams
+
+    async with build_async_engine_client_from_engine_args(
+        engine_args,
+        disable_frontend_multiprocessing=disable_frontend_multiprocessing,
+    ) as llm:
+        model_config = await llm.get_model_config()
+        assert all(
+            model_config.max_model_len
+            >= (request.prompt_len + request.expected_output_len)
+            for request in requests
+        ), (
+            "Please ensure that max_model_len is greater than the sum of"
+            " prompt_len and expected_output_len for all requests."
+        )
+
+        # Add the requests to the engine.
+        prompts: list[Union[TextPrompt, TokensPrompt]] = []
+        sampling_params: list[SamplingParams] = []
+        lora_requests: list[Optional[LoRARequest]] = []
+        for request in requests:
+            prompts.append(
+                TokensPrompt(
+                    prompt_token_ids=request.prompt["prompt_token_ids"],
+                    multi_modal_data=request.multi_modal_data,
+                )
+                if "prompt_token_ids" in request.prompt
+                else TextPrompt(
+                    prompt=request.prompt, multi_modal_data=request.multi_modal_data
+                )
+            )
+            sampling_params.append(
+                SamplingParams(
+                    n=n,
+                    temperature=1.0,
+                    top_p=1.0,
+                    ignore_eos=True,
+                    max_tokens=request.expected_output_len,
+                    detokenize=not disable_detokenize,
+                )
+            )
+            lora_requests.append(request.lora_request)
+
+        generators = []
+        start = time.perf_counter()
+        for i, (prompt, sp, lr) in enumerate(
+            zip(prompts, sampling_params, lora_requests)
+        ):
+            generator = llm.generate(prompt, sp, lora_request=lr, request_id=f"test{i}")
+            generators.append(generator)
+        all_gens = merge_async_iterators(*generators)
+        async for i, res in all_gens:
+            pass
+        end = time.perf_counter()
+        return end - start
+
+
+def run_hf(
+    requests: list[SampleRequest],
+    model: str,
+    tokenizer: PreTrainedTokenizerBase,
+    n: int,
+    max_batch_size: int,
+    trust_remote_code: bool,
+    disable_detokenize: bool = False,
+) -> float:
+    llm = AutoModelForCausalLM.from_pretrained(
+        model, torch_dtype=torch.float16, trust_remote_code=trust_remote_code
+    )
+    if llm.config.model_type == "llama":
+        # To enable padding in the HF backend.
+        tokenizer.pad_token = tokenizer.eos_token
+    llm = llm.cuda()
+
+    pbar = tqdm(total=len(requests))
+    start = time.perf_counter()
+    batch: list[str] = []
+    max_prompt_len = 0
+    max_output_len = 0
+    for i in range(len(requests)):
+        prompt = requests[i].prompt
+        prompt_len = requests[i].prompt_len
+        output_len = requests[i].expected_output_len
+        # Add the prompt to the batch.
+        batch.append(prompt)
+        max_prompt_len = max(max_prompt_len, prompt_len)
+        max_output_len = max(max_output_len, output_len)
+        if len(batch) < max_batch_size and i != len(requests) - 1:
+            # Check if we can add more requests to the batch.
+            next_prompt_len = requests[i + 1].prompt_len
+            next_output_len = requests[i + 1].expected_output_len
+            if (
+                max(max_prompt_len, next_prompt_len)
+                + max(max_output_len, next_output_len)
+            ) <= 2048:
+                # We can add more requests to the batch.
+                continue
+
+        # Generate the sequences.
+        input_ids = tokenizer(batch, return_tensors="pt", padding=True).input_ids
+        llm_outputs = llm.generate(
+            input_ids=input_ids.cuda(),
+            do_sample=True,
+            num_return_sequences=n,
+            temperature=1.0,
+            top_p=1.0,
+            use_cache=True,
+            max_new_tokens=max_output_len,
+        )
+        if not disable_detokenize:
+            # Include the decoding time.
+            tokenizer.batch_decode(llm_outputs, skip_special_tokens=True)
+        pbar.update(len(batch))
+
+        # Clear the batch.
+        batch = []
+        max_prompt_len = 0
+        max_output_len = 0
+    end = time.perf_counter()
+    return end - start
+
+
+def run_mii(
+    requests: list[SampleRequest],
+    model: str,
+    tensor_parallel_size: int,
+    output_len: int,
+) -> float:
+    from mii import client, serve
+
+    llm = serve(model, tensor_parallel=tensor_parallel_size)
+    prompts = [request.prompt for request in requests]
+
+    start = time.perf_counter()
+    llm.generate(prompts, max_new_tokens=output_len)
+    end = time.perf_counter()
+    client = client(model)
+    client.terminate_server()
+    return end - start
+
+
+def save_to_pytorch_benchmark_format(
+    args: argparse.Namespace, results: dict[str, Any]
+) -> None:
+    pt_records = convert_to_pytorch_benchmark_format(
+        args=args,
+        metrics={
+            "requests_per_second": [results["requests_per_second"]],
+            "tokens_per_second": [results["tokens_per_second"]],
+        },
+        extra_info={
+            k: results[k] for k in ["elapsed_time", "num_requests", "total_num_tokens"]
+        },
+    )
+    if pt_records:
+        # Don't use json suffix here as we don't want CI to pick it up
+        pt_file = f"{os.path.splitext(args.output_json)[0]}.pytorch.json"
+        write_to_json(pt_file, pt_records)
+
+
+def get_requests(args, tokenizer):
+    # Common parameters for all dataset types.
+    common_kwargs = {
+        "dataset_path": args.dataset_path,
+        "random_seed": args.seed,
+    }
+    sample_kwargs = {
+        "tokenizer": tokenizer,
+        "lora_path": args.lora_path,
+        "max_loras": args.max_loras,
+        "num_requests": args.num_prompts,
+        "input_len": args.input_len,
+        "output_len": args.output_len,
+    }
+
+    if args.dataset_path is None or args.dataset_name == "random":
+        sample_kwargs["range_ratio"] = args.random_range_ratio
+        sample_kwargs["prefix_len"] = args.prefix_len
+        dataset_cls = RandomDataset
+    elif args.dataset_name == "sharegpt":
+        dataset_cls = ShareGPTDataset
+        if args.backend == "vllm-chat":
+            sample_kwargs["enable_multimodal_chat"] = True
+    elif args.dataset_name == "sonnet":
+        assert tokenizer.chat_template or tokenizer.default_chat_template, (
+            "Tokenizer/model must have chat template for sonnet dataset."
+        )
+        dataset_cls = SonnetDataset
+        sample_kwargs["prefix_len"] = args.prefix_len
+        sample_kwargs["return_prompt_formatted"] = True
+    elif args.dataset_name == "burstgpt":
+        dataset_cls = BurstGPTDataset
+    elif args.dataset_name == "hf":
+        common_kwargs["no_stream"] = args.no_stream
+        if args.dataset_path in VisionArenaDataset.SUPPORTED_DATASET_PATHS:
+            dataset_cls = VisionArenaDataset
+            common_kwargs["dataset_subset"] = None
+            common_kwargs["dataset_split"] = "train"
+            sample_kwargs["enable_multimodal_chat"] = True
+        elif args.dataset_path in InstructCoderDataset.SUPPORTED_DATASET_PATHS:
+            dataset_cls = InstructCoderDataset
+            common_kwargs["dataset_split"] = "train"
+        elif args.dataset_path in ConversationDataset.SUPPORTED_DATASET_PATHS:
+            dataset_cls = ConversationDataset
+            common_kwargs["dataset_subset"] = args.hf_subset
+            common_kwargs["dataset_split"] = args.hf_split
+            sample_kwargs["enable_multimodal_chat"] = True
+        elif args.dataset_path in AIMODataset.SUPPORTED_DATASET_PATHS:
+            dataset_cls = AIMODataset
+            common_kwargs["dataset_subset"] = None
+            common_kwargs["dataset_split"] = "train"
+    else:
+        raise ValueError(f"Unknown dataset name: {args.dataset_name}")
+    # Remove None values
+    sample_kwargs = {k: v for k, v in sample_kwargs.items() if v is not None}
+    return dataset_cls(**common_kwargs).sample(**sample_kwargs)
+
+
+@deprecated(
+    "benchmark_throughput.py is deprecated and will be removed in a "
+    "future version. Please use 'vllm bench throughput' instead.",
+)
+def main(args: argparse.Namespace):
+    if args.seed is None:
+        args.seed = 0
+    print(args)
+    random.seed(args.seed)
+    # Sample the requests.
+    tokenizer = AutoTokenizer.from_pretrained(
+        args.tokenizer, trust_remote_code=args.trust_remote_code
+    )
+    requests = get_requests(args, tokenizer)
+    is_multi_modal = any(request.multi_modal_data is not None for request in requests)
+    request_outputs: Optional[list[RequestOutput]] = None
+    if args.backend == "vllm":
+        if args.async_engine:
+            elapsed_time = uvloop.run(
+                run_vllm_async(
+                    requests,
+                    args.n,
+                    AsyncEngineArgs.from_cli_args(args),
+                    args.disable_frontend_multiprocessing,
+                    args.disable_detokenize,
+                )
+            )
+        else:
+            elapsed_time, request_outputs = run_vllm(
+                requests,
+                args.n,
+                EngineArgs.from_cli_args(args),
+                args.disable_detokenize,
+            )
+    elif args.backend == "hf":
+        assert args.tensor_parallel_size == 1
+        elapsed_time = run_hf(
+            requests,
+            args.model,
+            tokenizer,
+            args.n,
+            args.hf_max_batch_size,
+            args.trust_remote_code,
+            args.disable_detokenize,
+        )
+    elif args.backend == "mii":
+        elapsed_time = run_mii(
+            requests, args.model, args.tensor_parallel_size, args.output_len
+        )
+    elif args.backend == "vllm-chat":
+        elapsed_time, request_outputs = run_vllm_chat(
+            requests, args.n, EngineArgs.from_cli_args(args), args.disable_detokenize
+        )
+    else:
+        raise ValueError(f"Unknown backend: {args.backend}")
+
+    if request_outputs:
+        # Note: with the vllm and vllm-chat backends,
+        # we have request_outputs, which we use to count tokens.
+        total_prompt_tokens = 0
+        total_output_tokens = 0
+        for ro in request_outputs:
+            if not isinstance(ro, RequestOutput):
+                continue
+            total_prompt_tokens += (
+                len(ro.prompt_token_ids) if ro.prompt_token_ids else 0
+            )
+            total_output_tokens += sum(len(o.token_ids) for o in ro.outputs if o)
+        total_num_tokens = total_prompt_tokens + total_output_tokens
+    else:
+        total_num_tokens = sum(r.prompt_len + r.expected_output_len for r in requests)
+        total_output_tokens = sum(r.expected_output_len for r in requests)
+        total_prompt_tokens = total_num_tokens - total_output_tokens
+
+    if is_multi_modal and args.backend != "vllm-chat":
+        print(
+            "\033[91mWARNING\033[0m: Multi-modal request with "
+            f"{args.backend} backend detected. The "
+            "following metrics are not accurate because image tokens are not"
+            " counted. See vllm-project/vllm/issues/9778 for details."
+        )
+        # TODO(vllm-project/vllm/issues/9778): Count multi-modal token length.
+        # vllm-chat backend counts the image tokens now
+
+    print(
+        f"Throughput: {len(requests) / elapsed_time:.2f} requests/s, "
+        f"{total_num_tokens / elapsed_time:.2f} total tokens/s, "
+        f"{total_output_tokens / elapsed_time:.2f} output tokens/s"
+    )
+    print(f"Total num prompt tokens:  {total_prompt_tokens}")
+    print(f"Total num output tokens:  {total_output_tokens}")
+
+    # Output JSON results if specified
+    if args.output_json:
+        results = {
+            "elapsed_time": elapsed_time,
+            "num_requests": len(requests),
+            "total_num_tokens": total_num_tokens,
+            "requests_per_second": len(requests) / elapsed_time,
+            "tokens_per_second": total_num_tokens / elapsed_time,
+        }
+        with open(args.output_json, "w") as f:
+            json.dump(results, f, indent=4)
+        save_to_pytorch_benchmark_format(args, results)
+
+
+def validate_args(args):
+    """
+    Validate command-line arguments.
+    """
+
+    # === Deprecation and Defaulting ===
+    if args.dataset is not None:
+        warnings.warn(
+            "The '--dataset' argument will be deprecated in the next release. "
+            "Please use '--dataset-name' and '--dataset-path' instead.",
+            stacklevel=2,
+        )
+        args.dataset_path = args.dataset
+
+    if not getattr(args, "tokenizer", None):
+        args.tokenizer = args.model
+
+    # === Backend Validation ===
+    valid_backends = {"vllm", "hf", "mii", "vllm-chat"}
+    if args.backend not in valid_backends:
+        raise ValueError(f"Unsupported backend: {args.backend}")
+
+    # === Dataset Configuration ===
+    if not args.dataset and not args.dataset_path:
+        print("When dataset path is not set, it will default to random dataset")
+        args.dataset_name = "random"
+        if args.input_len is None:
+            raise ValueError("input_len must be provided for a random dataset")
+
+    # === Dataset Name Specific Checks ===
+    # --hf-subset and --hf-split: only used
+    # when dataset_name is 'hf'
+    if args.dataset_name != "hf" and (
+        getattr(args, "hf_subset", None) is not None
+        or getattr(args, "hf_split", None) is not None
+    ):
+        warnings.warn(
+            "--hf-subset and --hf-split will be ignored \
+                since --dataset-name is not 'hf'.",
+            stacklevel=2,
+        )
+    elif args.dataset_name == "hf":
+        if args.dataset_path in (
+            VisionArenaDataset.SUPPORTED_DATASET_PATHS.keys()
+            | ConversationDataset.SUPPORTED_DATASET_PATHS
+        ):
+            assert args.backend == "vllm-chat", (
+                f"{args.dataset_path} needs to use vllm-chat as the backend."
+            )  # noqa: E501
+        elif args.dataset_path in (
+            InstructCoderDataset.SUPPORTED_DATASET_PATHS
+            | AIMODataset.SUPPORTED_DATASET_PATHS
+        ):
+            assert args.backend == "vllm", (
+                f"{args.dataset_path} needs to use vllm as the backend."
+            )  # noqa: E501
+        else:
+            raise ValueError(f"{args.dataset_path} is not supported by hf dataset.")
+
+    # --random-range-ratio: only used when dataset_name is 'random'
+    if args.dataset_name != "random" and args.random_range_ratio is not None:
+        warnings.warn(
+            "--random-range-ratio will be ignored since \
+                --dataset-name is not 'random'.",
+            stacklevel=2,
+        )
+
+    # --prefix-len: only used when dataset_name is 'random', 'sonnet', or not
+    # set.
+    if (
+        args.dataset_name not in {"random", "sonnet", None}
+        and args.prefix_len is not None
+    ):
+        warnings.warn(
+            "--prefix-len will be ignored since --dataset-name\
+                 is not 'random', 'sonnet', or not set.",
+            stacklevel=2,
+        )
+
+    # === LoRA Settings ===
+    if getattr(args, "enable_lora", False) and args.backend != "vllm":
+        raise ValueError("LoRA benchmarking is only supported for vLLM backend")
+    if getattr(args, "enable_lora", False) and args.lora_path is None:
+        raise ValueError("LoRA path must be provided when enable_lora is True")
+
+    # === Backend-specific Validations ===
+    if args.backend == "hf" and args.hf_max_batch_size is None:
+        raise ValueError("HF max batch size is required for HF backend")
+    if args.backend != "hf" and args.hf_max_batch_size is not None:
+        raise ValueError("HF max batch size is only for HF backend.")
+
+    if (
+        args.backend in {"hf", "mii"}
+        and getattr(args, "quantization", None) is not None
+    ):
+        raise ValueError("Quantization is only for vLLM backend.")
+
+    if args.backend == "mii" and args.dtype != "auto":
+        raise ValueError("dtype must be auto for MII backend.")
+    if args.backend == "mii" and args.n != 1:
+        raise ValueError("n must be 1 for MII backend.")
+    if args.backend == "mii" and args.tokenizer != args.model:
+        raise ValueError("Tokenizer must be the same as the model for MII backend.")
+
+    # --data-parallel is not supported currently.
+    # https://github.com/vllm-project/vllm/issues/16222
+    if args.data_parallel_size > 1:
+        raise ValueError(
+            "Data parallel is not supported in offline benchmark, "
+            "please use benchmark serving instead"
+        )
+
+
+def create_argument_parser():
+    parser = FlexibleArgumentParser(description="Benchmark the throughput.")
+    parser.add_argument(
+        "--backend",
+        type=str,
+        choices=["vllm", "hf", "mii", "vllm-chat"],
+        default="vllm",
+    )
+    parser.add_argument(
+        "--dataset-name",
+        type=str,
+        choices=["sharegpt", "random", "sonnet", "burstgpt", "hf"],
+        help="Name of the dataset to benchmark on.",
+        default="sharegpt",
+    )
+    parser.add_argument(
+        "--no-stream",
+        action="store_true",
+        help="Do not load the dataset in streaming mode.",
+    )
+    parser.add_argument(
+        "--dataset",
+        type=str,
+        default=None,
+        help="Path to the ShareGPT dataset, will be deprecated in\
+            the next release. The dataset is expected to "
+        "be a json in form of list[dict[..., conversations: "
+        "list[dict[..., value: <prompt_or_response>]]]]",
+    )
+    parser.add_argument(
+        "--dataset-path", type=str, default=None, help="Path to the dataset"
+    )
+    parser.add_argument(
+        "--input-len",
+        type=int,
+        default=None,
+        help="Input prompt length for each request",
+    )
+    parser.add_argument(
+        "--output-len",
+        type=int,
+        default=None,
+        help="Output length for each request. Overrides the "
+        "output length from the dataset.",
+    )
+    parser.add_argument(
+        "--n", type=int, default=1, help="Number of generated sequences per prompt."
+    )
+    parser.add_argument(
+        "--num-prompts", type=int, default=1000, help="Number of prompts to process."
+    )
+    parser.add_argument(
+        "--hf-max-batch-size",
+        type=int,
+        default=None,
+        help="Maximum batch size for HF backend.",
+    )
+    parser.add_argument(
+        "--output-json",
+        type=str,
+        default=None,
+        help="Path to save the throughput results in JSON format.",
+    )
+    parser.add_argument(
+        "--async-engine",
+        action="store_true",
+        default=False,
+        help="Use vLLM async engine rather than LLM class.",
+    )
+    parser.add_argument(
+        "--disable-frontend-multiprocessing",
+        action="store_true",
+        default=False,
+        help="Disable decoupled async engine frontend.",
+    )
+    parser.add_argument(
+        "--disable-detokenize",
+        action="store_true",
+        help=(
+            "Do not detokenize the response (i.e. do not include "
+            "detokenization time in the measurement)"
+        ),
+    )
+    # LoRA
+    parser.add_argument(
+        "--lora-path",
+        type=str,
+        default=None,
+        help="Path to the LoRA adapters to use. This can be an absolute path, "
+        "a relative path, or a Hugging Face model identifier.",
+    )
+    parser.add_argument(
+        "--prefix-len",
+        type=int,
+        default=None,
+        help=f"Number of prefix tokens to be used in RandomDataset "
+        "and SonnetDataset. For RandomDataset, the total input "
+        "length is the sum of prefix-len (default: "
+        f"{RandomDataset.DEFAULT_PREFIX_LEN}) and a random context length "
+        "sampled from [input_len * (1 - range_ratio), "
+        "input_len * (1 + range_ratio)]. For SonnetDataset, "
+        f"prefix_len (default: {SonnetDataset.DEFAULT_PREFIX_LEN}) "
+        "controls how much of the input is fixed lines versus "
+        "random lines, but the total input length remains approximately "
+        "input_len tokens.",
+    )
+    # random dataset
+    parser.add_argument(
+        "--random-range-ratio",
+        type=float,
+        default=None,
+        help=f"Range ratio (default : {RandomDataset.DEFAULT_RANGE_RATIO}) "
+        "for sampling input/output length, "
+        "used only for RandomDataset. Must be in the range [0, 1) to "
+        "define a symmetric sampling range "
+        "[length * (1 - range_ratio), length * (1 + range_ratio)].",
+    )
+
+    # hf dataset
+    parser.add_argument(
+        "--hf-subset", type=str, default=None, help="Subset of the HF dataset."
+    )
+    parser.add_argument(
+        "--hf-split", type=str, default=None, help="Split of the HF dataset."
+    )
+
+    parser = AsyncEngineArgs.add_cli_args(parser)
+
+    return parser
+
 
 if __name__ == "__main__":
-    print("""DEPRECATED: This script has been moved to the vLLM CLI.
-
-Please use the following command instead:
-    vllm bench throughput
-
-For help with the new command, run:
-    vllm bench throughput --help
-
-Alternatively, you can run the new command directly with:
-    python -m vllm.entrypoints.cli.main bench throughput --help
-""")
-    sys.exit(1)
+    parser = create_argument_parser()
+    args = parser.parse_args()
+    if args.tokenizer is None:
+        args.tokenizer = args.model
+    validate_args(args)
+    main(args)

benchmarks/benchmark_utils.py

@@ -6,7 +6,7 @@ import math
 import os
 import time
 from types import TracebackType
-from typing import Any
+from typing import Any, Optional, Union
 
 
 def convert_to_pytorch_benchmark_format(
@@ -92,7 +92,7 @@ class TimeCollector:
     def __init__(self, scale: int) -> None:
         self.cnt: int = 0
         self._sum: int = 0
-        self._max: int | None = None
+        self._max: Optional[int] = None
         self.scale = scale
         self.start_time: int = time.monotonic_ns()
 
@@ -104,13 +104,13 @@ class TimeCollector:
         else:
             self._max = max(self._max, v)
 
-    def avg(self) -> float | str:
+    def avg(self) -> Union[float, str]:
         return self._sum * 1.0 / self.cnt / self.scale if self.cnt > 0 else "N/A"
 
-    def max(self) -> float | str:
+    def max(self) -> Union[float, str]:
         return self._max / self.scale if self._max else "N/A"
 
-    def dump_avg_max(self) -> list[float | str]:
+    def dump_avg_max(self) -> list[Union[float, str]]:
         return [self.avg(), self.max()]
 
     def __enter__(self) -> None:
@@ -118,8 +118,8 @@ class TimeCollector:
 
     def __exit__(
         self,
-        exc_type: type[BaseException] | None,
-        exc_value: BaseException | None,
-        exc_traceback: TracebackType | None,
+        exc_type: Optional[type[BaseException]],
+        exc_value: Optional[BaseException],
+        exc_traceback: Optional[TracebackType],
     ) -> None:
         self.collect(time.monotonic_ns() - self.start_time)

benchmarks/cutlass_benchmarks/sparse_benchmarks.py

@@ -6,7 +6,8 @@ import copy
 import itertools
 import pickle as pkl
 import time
-from collections.abc import Callable, Iterable
+from collections.abc import Iterable
+from typing import Callable
 
 import torch
 import torch.utils.benchmark as TBenchmark
@@ -15,7 +16,7 @@ from utils import make_rand_sparse_tensors
 from weight_shapes import WEIGHT_SHAPES
 
 from vllm import _custom_ops as ops
-from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils import FlexibleArgumentParser
 
 DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
 DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512]

benchmarks/cutlass_benchmarks/w8a8_benchmarks.py

@@ -6,7 +6,8 @@ import copy
 import itertools
 import pickle as pkl
 import time
-from collections.abc import Callable, Iterable
+from collections.abc import Iterable
+from typing import Callable, Optional
 
 import torch
 import torch.utils.benchmark as TBenchmark
@@ -16,10 +17,9 @@ from weight_shapes import WEIGHT_SHAPES
 
 from vllm import _custom_ops as ops
 from vllm.model_executor.layers.quantization.utils.fp8_utils import (
-    w8a8_triton_block_scaled_mm,
+    w8a8_block_fp8_matmul,
 )
-from vllm.utils.argparse_utils import FlexibleArgumentParser
-from vllm.utils.math_utils import cdiv
+from vllm.utils import FlexibleArgumentParser, cdiv
 
 DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
 DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512]
@@ -53,7 +53,7 @@ def bench_int8(
     n: int,
     label: str,
     sub_label: str,
-    bench_kernels: list[str] | None = None,
+    bench_kernels: Optional[list[str]] = None,
 ) -> Iterable[TMeasurement]:
     """Benchmark INT8-based kernels."""
     assert dtype == torch.int8
@@ -108,7 +108,7 @@ def bench_fp8(
     n: int,
     label: str,
     sub_label: str,
-    bench_kernels: list[str] | None = None,
+    bench_kernels: Optional[list[str]] = None,
 ) -> Iterable[TMeasurement]:
     """Benchmark FP8-based kernels."""
     assert dtype == torch.float8_e4m3fn
@@ -158,7 +158,7 @@ def bench_fp8(
         "cutlass_fp8_fp8_fp16_scaled_mm_bias": lambda: ops.cutlass_scaled_mm(
             a, b, scale_a, scale_b, torch.float16, bias.to(dtype=torch.float16)
         ),
-        "triton_fp8_fp8_fp16_scaled_mm_blockwise": lambda: w8a8_triton_block_scaled_mm(
+        "triton_fp8_fp8_fp16_scaled_mm_blockwise": lambda: w8a8_block_fp8_matmul(
             a_cont, b.t(), block_scale_a, block_scale_b.t(), (128, 128)
         ),
         "cutlass_fp8_fp8_fp16_scaled_mm_blockwise": lambda: ops.cutlass_scaled_mm(
@@ -183,7 +183,7 @@ def bench(
     n: int,
     label: str,
     sub_label: str,
-    bench_kernels: list[str] | None = None,
+    bench_kernels: Optional[list[str]] = None,
 ) -> Iterable[TMeasurement]:
     if dtype == torch.int8:
         return bench_int8(dtype, m, k, n, label, sub_label, bench_kernels)
@@ -201,7 +201,7 @@ def print_timers(timers: Iterable[TMeasurement]):
 def run(
     dtype: torch.dtype,
     MKNs: Iterable[tuple[int, int, int]],
-    bench_kernels: list[str] | None = None,
+    bench_kernels: Optional[list[str]] = None,
 ) -> Iterable[TMeasurement]:
     results = []
     for m, k, n in MKNs:

benchmarks/disagg_benchmarks/disagg_overhead_benchmark.sh

@@ -55,7 +55,9 @@ benchmark() {
   output_len=$2
 
 
-  CUDA_VISIBLE_DEVICES=0 vllm serve $model \
+  CUDA_VISIBLE_DEVICES=0 python3 \
+    -m vllm.entrypoints.openai.api_server \
+    --model $model \
     --port 8100 \
     --max-model-len 10000 \
     --gpu-memory-utilization 0.6 \
@@ -63,7 +65,9 @@ benchmark() {
     '{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2,"kv_buffer_size":5e9}' &
 
 
-  CUDA_VISIBLE_DEVICES=1 vllm serve $model \
+  CUDA_VISIBLE_DEVICES=1 python3 \
+    -m vllm.entrypoints.openai.api_server \
+    --model $model \
     --port 8200 \
     --max-model-len 10000 \
     --gpu-memory-utilization 0.6 \

benchmarks/disagg_benchmarks/disagg_performance_benchmark.sh

@@ -38,12 +38,16 @@ wait_for_server() {
 launch_chunked_prefill() {
   model="meta-llama/Meta-Llama-3.1-8B-Instruct"
   # disagg prefill
-  CUDA_VISIBLE_DEVICES=0 vllm serve $model \
+  CUDA_VISIBLE_DEVICES=0 python3 \
+    -m vllm.entrypoints.openai.api_server \
+    --model $model \
     --port 8100 \
     --max-model-len 10000 \
     --enable-chunked-prefill \
     --gpu-memory-utilization 0.6 &
-  CUDA_VISIBLE_DEVICES=1 vllm serve $model \
+  CUDA_VISIBLE_DEVICES=1 python3 \
+    -m vllm.entrypoints.openai.api_server \
+    --model $model \
     --port 8200 \
     --max-model-len 10000 \
     --enable-chunked-prefill \
@@ -58,14 +62,18 @@ launch_chunked_prefill() {
 launch_disagg_prefill() {
   model="meta-llama/Meta-Llama-3.1-8B-Instruct"
   # disagg prefill
-  CUDA_VISIBLE_DEVICES=0 vllm serve $model \
+  CUDA_VISIBLE_DEVICES=0 python3 \
+    -m vllm.entrypoints.openai.api_server \
+    --model $model \
     --port 8100 \
     --max-model-len 10000 \
     --gpu-memory-utilization 0.6 \
     --kv-transfer-config \
     '{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2,"kv_buffer_size":5e9}' &
 
-  CUDA_VISIBLE_DEVICES=1 vllm serve $model \
+  CUDA_VISIBLE_DEVICES=1 python3 \
+    -m vllm.entrypoints.openai.api_server \
+    --model $model \
     --port 8200 \
     --max-model-len 10000 \
     --gpu-memory-utilization 0.6 \

benchmarks/fused_kernels/layernorm_rms_benchmarks.py

@@ -3,9 +3,10 @@
 
 import pickle as pkl
 import time
-from collections.abc import Callable, Iterable
+from collections.abc import Iterable
 from dataclasses import dataclass
 from itertools import product
+from typing import Callable, Optional
 
 import torch
 import torch.utils.benchmark as TBenchmark
@@ -50,7 +51,7 @@ def get_bench_params() -> list[bench_params_t]:
 def unfused_int8_impl(
     rms_norm_layer: RMSNorm,
     x: torch.Tensor,
-    residual: torch.Tensor | None,
+    residual: Optional[torch.Tensor],
     quant_dtype: torch.dtype,
 ):
     # Norm
@@ -67,7 +68,7 @@ def unfused_int8_impl(
 def unfused_fp8_impl(
     rms_norm_layer: RMSNorm,
     x: torch.Tensor,
-    residual: torch.Tensor | None,
+    residual: Optional[torch.Tensor],
     quant_dtype: torch.dtype,
 ):
     # Norm
@@ -84,7 +85,7 @@ def unfused_fp8_impl(
 def fused_impl(
     rms_norm_layer: RMSNorm,  # this stores the weights
     x: torch.Tensor,
-    residual: torch.Tensor | None,
+    residual: Optional[torch.Tensor],
     quant_dtype: torch.dtype,
 ):
     out, _ = ops.rms_norm_dynamic_per_token_quant(