[Model] Use the same fused_moe configs for all H200 devices (#23642)

Signed-off-by: Roger Meier <r.meier@siemens.com>

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

@@ -0,0 +1,14 @@
+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-h100.txt

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

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

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

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

@@ -21,10 +21,13 @@ def launch_lm_eval(eval_config, tp_size):
     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=true,"
+        f"enforce_eager={enforce_eager},"
+        f"kv_cache_dtype={kv_cache_dtype},"
         f"add_bos_token=true,"
         f"trust_remote_code={trust_remote_code},"
         f"max_model_len={max_model_len},"
@@ -37,8 +40,13 @@ def launch_lm_eval(eval_config, tp_size):
         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.
-        apply_chat_template=backend == "vllm-vlm",
+        # 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,
     )
     return results

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

@@ -1,184 +0,0 @@
-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

@@ -1,28 +0,0 @@
-# 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

@@ -1,39 +0,0 @@
-
-# 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 guide: [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

@@ -1,196 +0,0 @@
-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/download-tokenizer.py

@@ -1,26 +0,0 @@
-# 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

@@ -1,97 +0,0 @@
-# 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

@@ -1,9 +0,0 @@
-# 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/nightly-annotate.sh

@@ -1,78 +0,0 @@
-#!/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

@@ -1,464 +0,0 @@
-#!/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/summary-nightly-results.py

@@ -1,82 +0,0 @@
-# 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

@@ -1,23 +0,0 @@
-#!/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/performance-benchmarks/README.md

@@ -2,40 +2,23 @@
 
 ## Introduction
 
-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.
+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.
 
 ## Performance benchmark quick overview
 
-**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 Coverage**: latency, throughput and fix-qps serving on B200, A100, H100 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
 
-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
+The benchmark needs to be triggered manually:
 
 ```bash
-bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+bash .buildkite/performance-benchmarks/scripts/run-performance-benchmarks.sh
 ```
 
 Runtime environment variables:
@@ -47,10 +30,6 @@ 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.
@@ -152,26 +131,3 @@ 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/performance-benchmarks/scripts/convert-results-json-to-markdown.py

@@ -392,7 +392,7 @@ if __name__ == "__main__":
     json_file = "benchmark_results.json"
     with open(results_folder / md_file, "w") as f:
         results = read_markdown(
-            "../.buildkite/nightly-benchmarks/"
+            "../.buildkite/performance-benchmarks/"
             + "performance-benchmarks-descriptions.md"
         )
         results = results.format(

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

@@ -469,7 +469,7 @@ main() {
   ensure_sharegpt_downloaded
   declare -g RESULTS_FOLDER=results/
   mkdir -p $RESULTS_FOLDER
-  QUICK_BENCHMARK_ROOT=../.buildkite/nightly-benchmarks/
+  QUICK_BENCHMARK_ROOT=../.buildkite/performance-benchmarks/
 
   # dump vllm info via vllm collect-env
   env_output=$(vllm collect-env)

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

@@ -20,7 +20,10 @@ trap remove_docker_container EXIT
 
 # Run the image and test offline inference/tensor parallel
 docker run \
-    --device /dev/dri \
+    --device /dev/dri:/dev/dri \
+    --net=host \
+    --ipc=host \
+    --privileged \
     -v /dev/dri/by-path:/dev/dri/by-path \
     --entrypoint="" \
     -e "HF_TOKEN=${HF_TOKEN}" \
@@ -42,7 +45,7 @@ docker run \
     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
+    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/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
 '

.buildkite/test-amd.yaml

@@ -38,7 +38,7 @@ 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/generate_nightly_torch_test.py
+  # in /vllm/tools/pre_commit/generate_nightly_torch_test.py
   mirror_hardwares: [amdexperimental]
   agent_pool: mi325_1
   # grade: Blocking
@@ -286,7 +286,7 @@ steps:
 
 - label: Engine Test # 25min
   timeout_in_minutes: 40
-  mirror_hardwares: [amdexperimental]
+  mirror_hardwares: [amdexperimental, amdproduction]
   agent_pool: mi325_1
   #grade: Blocking
   source_file_dependencies:
@@ -318,7 +318,7 @@ steps:
 
 - label: V1 Test entrypoints # 35min
   timeout_in_minutes: 50
-  mirror_hardwares: [amdexperimental]
+  mirror_hardwares: [amdexperimental, amdproduction]
   agent_pool: mi325_1
   # grade: Blocking
   source_file_dependencies:
@@ -561,7 +561,7 @@ steps:
 
 - label: Model Executor Test # 23min
   timeout_in_minutes: 35
-  mirror_hardwares: [amdexperimental]
+  mirror_hardwares: [amdexperimental, amdproduction]
   agent_pool: mi325_1
   # grade: Blocking
   source_file_dependencies:
@@ -789,8 +789,10 @@ steps:
   - vllm/
   - tests/models/language/generation
   commands:
-    # 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'
+    # 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'
     - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
 
 - label: Language Models Test (PPL)
@@ -906,7 +908,7 @@ steps:
 
 - label: Quantized Models Test # 45 min
   timeout_in_minutes: 60
-  mirror_hardwares: [amdexperimental]
+  mirror_hardwares: [amdexperimental, amdproduction]
   agent_pool: mi325_1
   # grade: Blocking
   source_file_dependencies:

.buildkite/test-pipeline.yaml

@@ -38,7 +38,7 @@ 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/generate_nightly_torch_test.py
+  # in /vllm/tools/pre_commit/generate_nightly_torch_test.py
   soft_fail: true
   source_file_dependencies:
   - requirements/nightly_torch_test.txt
@@ -205,6 +205,24 @@ 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"
@@ -313,6 +331,15 @@ steps:
     - 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/
@@ -392,7 +419,7 @@ steps:
       --ignore=lora/test_deepseekv2_tp.py \
       --ignore=lora/test_gptoss.py \
       --ignore=lora/test_qwen3moe_tp.py
-      
+
   parallelism: 4
 
 - label: PyTorch Compilation Unit Tests # 15min
@@ -435,6 +462,18 @@ steps:
   - 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
   mirror_hardwares: [amdexperimental]
@@ -477,6 +516,8 @@ 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
@@ -687,8 +728,10 @@ steps:
   - vllm/
   - tests/models/language/generation
   commands:
-    # 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'
+    # 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'
     - pytest -v -s models/language/generation -m '(not core_model) and (not hybrid_model)'
 
 - label: Language Models Test (PPL)
@@ -1101,7 +1144,7 @@ steps:
   - tests/weight_loading
   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"
@@ -1143,6 +1186,19 @@ 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
   gpu: h200

.github/mergify.yml

@@ -108,7 +108,7 @@ pull_request_rules:
       - files~=^benchmarks/
       - files~=^vllm/benchmarks/
       - files~=^tests/benchmarks/
-      - files~=^\.buildkite/nightly-benchmarks/
+      - files~=^\.buildkite/performance-benchmarks/
   actions:
     label:
       add:

.pre-commit-config.yaml

@@ -45,7 +45,7 @@ repos:
   - id: format-torch-nightly-test
     name: reformat nightly_torch_test.txt to be in sync with test.in
     language: python
-    entry: python tools/generate_nightly_torch_test.py
+    entry: python tools/pre_commit/generate_nightly_torch_test.py
     files: ^requirements/test\.(in|txt)$
   - id: mypy-local
     name: Run mypy locally for lowest supported Python version
@@ -78,12 +78,12 @@ repos:
     stages: [manual] # Only run in CI
   - id: shellcheck
     name: Lint shell scripts
-    entry: tools/shellcheck.sh
+    entry: tools/pre_commit/shellcheck.sh
     language: script
     types: [shell]
   - id: png-lint
     name: Lint PNG exports from excalidraw
-    entry: tools/png-lint.sh
+    entry: tools/pre_commit/png-lint.sh
     language: script
     types: [png]
   - id: signoff-commit
@@ -100,12 +100,12 @@ repos:
     stages: [commit-msg]
   - id: check-spdx-header
     name: Check SPDX headers
-    entry: python tools/check_spdx_header.py
+    entry: python tools/pre_commit/check_spdx_header.py
     language: python
     types: [python]
   - id: check-root-lazy-imports
     name: Check root lazy imports
-    entry: python tools/check_init_lazy_imports.py
+    entry: python tools/pre_commit/check_init_lazy_imports.py
     language: python
     types: [python]
   - id: check-filenames
@@ -119,11 +119,11 @@ repos:
     pass_filenames: false
   - id: update-dockerfile-graph
     name: Update Dockerfile dependency graph
-    entry: tools/update-dockerfile-graph.sh
+    entry: tools/pre_commit/update-dockerfile-graph.sh
     language: script
   - id: enforce-import-regex-instead-of-re
     name: Enforce import regex as re
-    entry: python tools/enforce_regex_import.py
+    entry: python tools/pre_commit/enforce_regex_import.py
     language: python
     types: [python]
     pass_filenames: false
@@ -131,7 +131,7 @@ repos:
   # forbid directly import triton
   - id: forbid-direct-triton-import
     name: "Forbid direct 'import triton'"
-    entry: python tools/check_triton_import.py
+    entry: python tools/pre_commit/check_triton_import.py
     language: python
     types: [python]
     pass_filenames: false
@@ -144,7 +144,7 @@ repos:
     additional_dependencies: [regex]
   - id: validate-config
     name: Validate configuration has default values and that each field has a docstring
-    entry: python tools/validate_config.py
+    entry: python tools/pre_commit/validate_config.py
     language: python
     additional_dependencies: [regex]
   # Keep `suggestion` last

README.md

@@ -21,6 +21,7 @@ Join us at the [PyTorch Conference, October 22-23](https://events.linuxfoundatio
 
 *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).

benchmarks/benchmark_block_pool.py

@@ -5,7 +5,7 @@ import gc
 from benchmark_utils import TimeCollector
 from tabulate import tabulate
 
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.v1.core.block_pool import BlockPool
 
 

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 import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 def test_long_document_qa(llm=None, sampling_params=None, prompts=None):

benchmarks/benchmark_ngram_proposer.py

@@ -19,7 +19,7 @@ from vllm.config import (
     VllmConfig,
 )
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_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

benchmarks/benchmark_prefix_caching.py

@@ -37,7 +37,7 @@ from transformers import PreTrainedTokenizerBase
 
 from vllm import LLM, SamplingParams
 from vllm.engine.arg_utils import EngineArgs
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 try:
     from vllm.transformers_utils.tokenizer import get_tokenizer

benchmarks/benchmark_prioritization.py

@@ -11,7 +11,7 @@ import time
 from transformers import AutoTokenizer, PreTrainedTokenizerBase
 
 from vllm.engine.arg_utils import EngineArgs
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 # Select a equi-probable random priority

benchmarks/benchmark_serving_structured_output.py

@@ -51,7 +51,7 @@ except ImportError:
     from backend_request_func import get_tokenizer
 
 try:
-    from vllm.utils import FlexibleArgumentParser
+    from vllm.utils.argparse_utils import FlexibleArgumentParser
 except ImportError:
     from argparse import ArgumentParser as FlexibleArgumentParser
 

benchmarks/cutlass_benchmarks/sparse_benchmarks.py

@@ -15,7 +15,7 @@ from utils import make_rand_sparse_tensors
 from weight_shapes import WEIGHT_SHAPES
 
 from vllm import _custom_ops as ops
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_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

@@ -18,7 +18,8 @@ from vllm import _custom_ops as ops
 from vllm.model_executor.layers.quantization.utils.fp8_utils import (
     w8a8_triton_block_scaled_mm,
 )
-from vllm.utils import FlexibleArgumentParser, cdiv
+from vllm.utils.argparse_utils import FlexibleArgumentParser
+from vllm.utils.math_utils import cdiv
 
 DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
 DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512]

benchmarks/kernels/bench_per_token_quant_fp8.py

@@ -10,7 +10,7 @@ import torch
 from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
 from vllm.model_executor.layers.quantization.utils.quant_utils import GroupShape
 from vllm.triton_utils import triton
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
 
 

benchmarks/kernels/benchmark_activation.py

@@ -10,7 +10,7 @@ import vllm.model_executor.layers.activation  # noqa F401
 from vllm.model_executor.custom_op import CustomOp
 from vllm.platforms import current_platform
 from vllm.triton_utils import triton
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
 
 batch_size_range = [1, 16, 32, 64, 128]

benchmarks/kernels/benchmark_bitblas.py

@@ -28,7 +28,7 @@ except ImportError as e:
 
 from bitblas import Matmul, MatmulConfig, auto_detect_nvidia_target
 
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 parser = FlexibleArgumentParser(
     description="Benchmark BitBLAS int4 on a specific target."

benchmarks/kernels/benchmark_cutlass_fp4_moe.py

@@ -20,7 +20,7 @@ from vllm.model_executor.layers.fused_moe.config import (
 from vllm.model_executor.layers.fused_moe.cutlass_moe import cutlass_moe_fp4
 from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts, fused_topk
 from vllm.scalar_type import scalar_types
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 WEIGHT_SHAPES_MOE = {
     "nvidia/DeepSeek-R1-FP4": [

benchmarks/kernels/benchmark_cutlass_moe_fp8.py

@@ -14,7 +14,7 @@ from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_confi
 from vllm.model_executor.layers.fused_moe.cutlass_moe import cutlass_moe_fp8
 from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts, fused_topk
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 # Weight shapes for different models: [num_experts, topk, hidden_size,
 # intermediate_size]

benchmarks/kernels/benchmark_device_communicators.py

@@ -39,7 +39,7 @@ from vllm.distributed.device_communicators.pynccl_allocator import (
 )
 from vllm.distributed.device_communicators.symm_mem import SymmMemCommunicator
 from vllm.logger import init_logger
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 logger = init_logger(__name__)
 

benchmarks/kernels/benchmark_grouped_gemm_cutlass.py

@@ -13,7 +13,7 @@ from vllm.model_executor.layers.fused_moe.fused_moe import (
     fused_experts,
     fused_topk,
 )
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 DEFAULT_MODELS = [
     "nm-testing/Mixtral-8x7B-Instruct-v0.1",

benchmarks/kernels/benchmark_layernorm.py

@@ -7,7 +7,7 @@ import torch
 
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
 
 

benchmarks/kernels/benchmark_lora.py

@@ -25,7 +25,7 @@ if HAS_TRITON:
     from vllm.lora.ops.triton_ops import LoRAKernelMeta, lora_expand, lora_shrink
     from vllm.lora.ops.triton_ops.utils import _LORA_A_PTR_DICT, _LORA_B_PTR_DICT
 
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 DEFAULT_MODELS = list(WEIGHT_SHAPES.keys())
 DEFAULT_TP_SIZES = [1]

benchmarks/kernels/benchmark_machete.py

@@ -33,7 +33,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
     quantize_weights,
 )
 from vllm.scalar_type import ScalarType, scalar_types
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 DEFAULT_MODELS = ["meta-llama/Llama-3-8b", "meta-llama/Llama-2-70b-hf"]
 DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512, 1024]

benchmarks/kernels/benchmark_marlin.py

@@ -44,7 +44,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
     sort_weights,
 )
 from vllm.scalar_type import ScalarType, scalar_types
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 DEFAULT_MODELS = ["meta-llama/Llama-2-7b-hf/TP1"]
 DEFAULT_BATCH_SIZES = [1, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192]

benchmarks/kernels/benchmark_moe.py

@@ -22,7 +22,7 @@ from vllm.model_executor.layers.fused_moe.fused_moe import *
 from vllm.platforms import current_platform
 from vllm.transformers_utils.config import get_config
 from vllm.triton_utils import triton
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 FP8_DTYPE = current_platform.fp8_dtype()
 

benchmarks/kernels/benchmark_moe_permute_unpermute.py

@@ -17,7 +17,7 @@ from vllm.model_executor.layers.fused_moe.moe_permute_unpermute import (
 )
 from vllm.model_executor.layers.fused_moe.utils import _fp8_quantize
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 FP8_DTYPE = current_platform.fp8_dtype()
 

benchmarks/kernels/benchmark_mrope.py

@@ -39,7 +39,7 @@ import torch
 from vllm.model_executor.layers.rotary_embedding import get_rope
 from vllm.platforms import current_platform
 from vllm.transformers_utils.config import get_config
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 

benchmarks/kernels/benchmark_paged_attention.py

@@ -9,7 +9,7 @@ import torch
 from vllm import _custom_ops as ops
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import (
     STR_DTYPE_TO_TORCH_DTYPE,
     create_kv_caches_with_random,

benchmarks/kernels/benchmark_quant.py

@@ -7,7 +7,7 @@ import torch
 
 from vllm import _custom_ops as ops
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import STR_DTYPE_TO_TORCH_DTYPE
 
 

benchmarks/kernels/benchmark_reshape_and_cache.py

@@ -9,7 +9,7 @@ from tabulate import tabulate
 from vllm import _custom_ops as ops
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import (
     STR_DTYPE_TO_TORCH_DTYPE,
     create_kv_caches_with_random,

benchmarks/kernels/benchmark_reshape_and_cache_flash.py

@@ -12,7 +12,7 @@ from vllm.attention.ops.triton_reshape_and_cache_flash import (
 )
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.torch_utils import (
     STR_DTYPE_TO_TORCH_DTYPE,
     create_kv_caches_with_random_flash,

benchmarks/kernels/benchmark_rope.py

@@ -8,7 +8,7 @@ import torch
 
 from vllm.model_executor.layers.rotary_embedding import RotaryEmbedding, get_rope
 from vllm.platforms import current_platform
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 def benchmark_rope_kernels_multi_lora(

benchmarks/kernels/benchmark_trtllm_decode_attention.py

@@ -8,7 +8,7 @@ from datetime import datetime
 import flashinfer
 import torch
 
-from vllm.utils import round_up
+from vllm.utils.math_utils import round_up
 
 FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
 FP8_DTYPE = torch.float8_e4m3fn

benchmarks/kernels/benchmark_trtllm_prefill_attention.py

@@ -8,7 +8,7 @@ from datetime import datetime
 import flashinfer
 import torch
 
-from vllm.utils import round_up
+from vllm.utils.math_utils import round_up
 
 FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
 FP8_DTYPE = torch.float8_e4m3fn

benchmarks/kernels/benchmark_w8a8_block_fp8.py

@@ -18,7 +18,7 @@ from vllm.model_executor.layers.quantization.utils.fp8_utils import (
 )
 from vllm.platforms import current_platform
 from vllm.triton_utils import triton
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 mp.set_start_method("spawn", force=True)
 

benchmarks/kernels/graph_machete_bench.py

@@ -11,7 +11,7 @@ import regex as re
 import seaborn as sns
 from torch.utils.benchmark import Measurement as TMeasurement
 
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 if __name__ == "__main__":
     parser = FlexibleArgumentParser(

benchmarks/overheads/benchmark_hashing.py

@@ -5,7 +5,7 @@ import cProfile
 import pstats
 
 from vllm import LLM, SamplingParams
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 # A very long prompt, total number of tokens is about 15k.
 LONG_PROMPT = ["You are an expert in large language models, aren't you?"] * 1000

cmake/cpu_extension.cmake

@@ -212,11 +212,24 @@ if ((AVX512_FOUND AND NOT AVX512_DISABLED) OR (ASIMD_FOUND AND NOT APPLE_SILICON
         # Build ACL with scons
         include(ProcessorCount)
         ProcessorCount(_NPROC)
+        set(_scons_cmd
+        scons -j${_NPROC}
+            Werror=0 debug=0 neon=1 examples=0 embed_kernels=0 os=linux
+            arch=armv8.2-a build=native benchmark_examples=0 fixed_format_kernels=1
+            multi_isa=1 openmp=1 cppthreads=0
+        )
+
+        # locate PyTorch's libgomp (e.g. site-packages/torch.libs/libgomp-947d5fa1.so.1.0.0)
+        # and create a local shim dir with it
+        include("${CMAKE_CURRENT_LIST_DIR}/utils.cmake")
+        vllm_prepare_torch_gomp_shim(VLLM_TORCH_GOMP_SHIM_DIR)
+
+        if(NOT VLLM_TORCH_GOMP_SHIM_DIR STREQUAL "")
+            list(APPEND _scons_cmd extra_link_flags=-L${VLLM_TORCH_GOMP_SHIM_DIR})
+        endif()
+
         execute_process(
-            COMMAND scons -j${_NPROC}
-                    Werror=0 debug=0 neon=1 examples=0 embed_kernels=0 os=linux
-                    arch=armv8.2-a build=native benchmark_examples=0 fixed_format_kernels=1
-                    multi_isa=1 openmp=1 cppthreads=0
+            COMMAND ${_scons_cmd}
             WORKING_DIRECTORY "$ENV{ACL_ROOT_DIR}"
             RESULT_VARIABLE _acl_rc
         )

cmake/utils.cmake

@@ -129,6 +129,44 @@ function (get_torch_gpu_compiler_flags OUT_GPU_FLAGS GPU_LANG)
   set(${OUT_GPU_FLAGS} ${GPU_FLAGS} PARENT_SCOPE)
 endfunction()
 
+# Find libgomp that gets shipped with PyTorch wheel and create a shim dir with:
+#   libgomp.so    -> libgomp-<hash>.so...
+#   libgomp.so.1  -> libgomp-<hash>.so...
+# OUTPUT: TORCH_GOMP_SHIM_DIR  ("" if not found)
+function(vllm_prepare_torch_gomp_shim TORCH_GOMP_SHIM_DIR)
+  set(${TORCH_GOMP_SHIM_DIR} "" PARENT_SCOPE)
+
+  # Use run_python to locate vendored libgomp; never throw on failure.
+  run_python(_VLLM_TORCH_GOMP_PATH
+    "
+import os, glob
+try:
+  import torch
+  torch_pkg = os.path.dirname(torch.__file__)
+  site_root = os.path.dirname(torch_pkg)
+  torch_libs = os.path.join(site_root, 'torch.libs')
+  print(glob.glob(os.path.join(torch_libs, 'libgomp-*.so*'))[0])
+except:
+  print('')
+"
+    "failed to probe torch.libs for libgomp")
+
+  if(_VLLM_TORCH_GOMP_PATH STREQUAL "" OR NOT EXISTS "${_VLLM_TORCH_GOMP_PATH}")
+    return()
+  endif()
+
+  # Create shim under the build tree
+  set(_shim "${CMAKE_BINARY_DIR}/gomp_shim")
+  file(MAKE_DIRECTORY "${_shim}")
+
+  execute_process(COMMAND ${CMAKE_COMMAND} -E rm -f "${_shim}/libgomp.so")
+  execute_process(COMMAND ${CMAKE_COMMAND} -E rm -f "${_shim}/libgomp.so.1")
+  execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink "${_VLLM_TORCH_GOMP_PATH}" "${_shim}/libgomp.so")
+  execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink "${_VLLM_TORCH_GOMP_PATH}" "${_shim}/libgomp.so.1")
+
+  set(${TORCH_GOMP_SHIM_DIR} "${_shim}" PARENT_SCOPE)
+endfunction()
+
 # Macro for converting a `gencode` version number to a cmake version number.
 macro(string_to_ver OUT_VER IN_STR)
   string(REGEX REPLACE "\([0-9]+\)\([0-9]\)" "\\1.\\2" ${OUT_VER} ${IN_STR})

docker/Dockerfile

@@ -361,13 +361,6 @@ RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist
     && uv pip install --system dist/*.whl --verbose \
         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
 
-# TODO (huydhn): Remove this once xformers is released for 2.9.0
-RUN --mount=type=cache,target=/root/.cache/uv bash - <<'BASH'
-    . /etc/environment
-    export TORCH_CUDA_ARCH_LIST='7.5 8.0+PTX 9.0a'
-    uv pip install --system --no-build-isolation "git+https://github.com/facebookresearch/xformers@v0.0.32.post2"
-BASH
-
 # Install FlashInfer pre-compiled kernel cache and binaries
 # https://docs.flashinfer.ai/installation.html
 RUN --mount=type=cache,target=/root/.cache/uv \
@@ -495,7 +488,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     else \
         BITSANDBYTES_VERSION="0.46.1"; \
     fi; \
-    uv pip install --system accelerate hf_transfer modelscope "bitsandbytes>=${BITSANDBYTES_VERSION}" 'timm>=1.0.17' 'runai-model-streamer[s3,gcs]>=0.14.0'
+    uv pip install --system accelerate hf_transfer modelscope "bitsandbytes>=${BITSANDBYTES_VERSION}" 'timm>=1.0.17' 'runai-model-streamer[s3,gcs]>=0.15.0'
 
 ENV VLLM_USAGE_SOURCE production-docker-image
 

docker/Dockerfile.cpu

@@ -79,7 +79,7 @@ RUN echo 'ulimit -c 0' >> ~/.bashrc
 ######################### BUILD IMAGE #########################
 FROM base AS vllm-build
 
-ARG max_jobs=2
+ARG max_jobs=32
 ENV MAX_JOBS=${max_jobs}
 
 ARG GIT_REPO_CHECK=0

docker/Dockerfile.rocm_base

@@ -7,7 +7,7 @@ ARG PYTORCH_REPO="https://github.com/ROCm/pytorch.git"
 ARG PYTORCH_VISION_REPO="https://github.com/pytorch/vision.git"
 ARG FA_BRANCH="0e60e394"
 ARG FA_REPO="https://github.com/Dao-AILab/flash-attention.git"
-ARG AITER_BRANCH="eef23c7f"
+ARG AITER_BRANCH="9716b1b8"
 ARG AITER_REPO="https://github.com/ROCm/aiter.git"
 
 FROM ${BASE_IMAGE} AS base

docs/cli/.nav.yml

@@ -5,4 +5,4 @@ nav:
   - complete.md
   - run-batch.md
   - vllm bench:
-    - bench/*.md
+    - bench/**/*.md

docs/cli/bench/sweep/plot.md

@@ -0,0 +1,9 @@
+# vllm bench sweep plot
+
+## JSON CLI Arguments
+
+--8<-- "docs/cli/json_tip.inc.md"
+
+## Options
+
+--8<-- "docs/argparse/bench_sweep_plot.md"

docs/cli/bench/sweep/serve.md

@@ -0,0 +1,9 @@
+# vllm bench sweep serve
+
+## JSON CLI Arguments
+
+--8<-- "docs/cli/json_tip.inc.md"
+
+## Options
+
+--8<-- "docs/argparse/bench_sweep_serve.md"

docs/cli/bench/sweep/serve_sla.md

@@ -0,0 +1,9 @@
+# vllm bench sweep serve_sla
+
+## JSON CLI Arguments
+
+--8<-- "docs/cli/json_tip.inc.md"
+
+## Options
+
+--8<-- "docs/argparse/bench_sweep_serve_sla.md"

docs/community/meetups.md

@@ -2,6 +2,7 @@
 
 We host regular meetups in San Francisco Bay Area every 2 months. We will share the project updates from the vLLM team and have guest speakers from the industry to share their experience and insights. Please find the materials of our previous meetups below:
 
+- [vLLM Shanghai Meetup](https://mp.weixin.qq.com/s/__xb4OyOsImz-9eAVrdlcg), October 25th 2025. [[Slides]](https://drive.google.com/drive/folders/1KqwjsFJLfEsC8wlDugnrR61zsWHt94Q6)
 - [vLLM Toronto Meetup](https://luma.com/e80e0ymm), September 25th 2025. [[Slides]](https://docs.google.com/presentation/d/1IYJYmJcu9fLpID5N5RbW_vO0XLo0CGOR14IXOjB61V8/edit?usp=sharing)
 - [vLLM Shenzhen Meetup](https://mp.weixin.qq.com/s/k8ZBO1u2_2odgiKWH_GVTQ), August 30th 2025. [[Slides]](https://drive.google.com/drive/folders/1Ua2SVKVSu-wp5vou_6ElraDt2bnKhiEA)
 - [vLLM Singapore Meetup](https://www.sginnovate.com/event/vllm-sg-meet), August 27th 2025. [[Slides]](https://drive.google.com/drive/folders/1ncf3GyqLdqFaB6IeB834E5TZJPLAOiXZ?usp=sharing)

docs/contributing/README.md

@@ -16,9 +16,9 @@ Finally, one of the most impactful ways to support us is by raising awareness ab
 Unsure on where to start? Check out the following links for tasks to work on:
 
 - [Good first issues](https://github.com/vllm-project/vllm/issues?q=is%3Aissue%20state%3Aopen%20label%3A%22good%20first%20issue%22)
-    - [Selected onboarding tasks](gh-project:6)
+    - [Selected onboarding tasks](https://github.com/orgs/vllm-project/projects/6)
 - [New model requests](https://github.com/vllm-project/vllm/issues?q=is%3Aissue%20state%3Aopen%20label%3A%22new-model%22)
-    - [Models with multi-modal capabilities](gh-project:10)
+    - [Models with multi-modal capabilities](https://github.com/orgs/vllm-project/projects/10)
 
 ## License
 

docs/contributing/benchmarks.md

@@ -9,7 +9,6 @@ vLLM provides comprehensive benchmarking tools for performance testing and evalu
 - **[Benchmark CLI](#benchmark-cli)**: `vllm bench` CLI tools and specialized benchmark scripts for interactive performance testing
 - **[Parameter sweeps](#parameter-sweeps)**: Automate `vllm bench` runs for multiple configurations
 - **[Performance benchmarks](#performance-benchmarks)**: Automated CI benchmarks for development
-- **[Nightly benchmarks](#nightly-benchmarks)**: Comparative benchmarks against alternatives
 
 [Benchmark CLI]: #benchmark-cli
 
@@ -321,6 +320,73 @@ The following arguments can be used to control the ramp-up:
 - `--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.
 
+##### Load Pattern Configuration
+
+vLLM's benchmark serving script provides sophisticated load pattern simulation capabilities through three key parameters that control request generation and concurrency behavior:
+
+###### Load Pattern Control Parameters
+
+- `--request-rate`: Controls the target request generation rate (requests per second). Set to `inf` for maximum throughput testing or finite values for controlled load simulation.
+- `--burstiness`: Controls traffic variability using a Gamma distribution (range: > 0). Lower values create bursty traffic, higher values create uniform traffic.
+- `--max-concurrency`: Limits concurrent outstanding requests. If this argument is not provided, concurrency is unlimited. Set a value to simulate backpressure.
+
+These parameters work together to create realistic load patterns with carefully chosen defaults. The `--request-rate` parameter defaults to `inf` (infinite), which sends all requests immediately for maximum throughput testing. When set to finite values, it uses either a Poisson process (default `--burstiness=1.0`) or Gamma distribution for realistic request timing. The `--burstiness` parameter only takes effect when `--request-rate` is not infinite - a value of 1.0 creates natural Poisson traffic, while lower values (0.1-0.5) create bursty patterns and higher values (2.0-5.0) create uniform spacing. The `--max-concurrency` parameter defaults to `None` (unlimited) but can be set to simulate real-world constraints where a load balancer or API gateway limits concurrent connections. When combined, these parameters allow you to simulate everything from unrestricted stress testing (`--request-rate=inf`) to production-like scenarios with realistic arrival patterns and resource constraints.
+
+The `--burstiness` parameter mathematically controls request arrival patterns using a Gamma distribution where:
+
+- Shape parameter: `burstiness` value
+- Coefficient of Variation (CV): $\frac{1}{\sqrt{burstiness}}$
+- Traffic characteristics:
+    - `burstiness = 0.1`: Highly bursty traffic (CV ≈ 3.16) - stress testing
+    - `burstiness = 1.0`: Natural Poisson traffic (CV = 1.0) - realistic simulation  
+    - `burstiness = 5.0`: Uniform traffic (CV ≈ 0.45) - controlled load testing
+
+![Load Pattern Examples](../assets/contributing/load-pattern-examples.png)
+
+*Figure: Load pattern examples for each use case. Top row: Request arrival timelines showing cumulative requests over time. Bottom row: Inter-arrival time distributions showing traffic variability patterns. Each column represents a different use case with its specific parameter settings and resulting traffic characteristics.*
+
+Load Pattern Recommendations by Use Case:
+
+| Use Case           | Burstiness   | Request Rate    | Max Concurrency | Description                                               |
+| ---                | ---          | ---             | ---             | ---                                                       |
+| Maximum Throughput | N/A          | Infinite        | Limited         | **Most common**: Simulates load balancer/gateway limits with unlimited user demand |
+| Realistic Testing  | 1.0          | Moderate (5-20) | Infinite        | Natural Poisson traffic patterns for baseline performance |
+| Stress Testing     | 0.1-0.5      | High (20-100)   | Infinite        | Challenging burst patterns to test resilience             |
+| Latency Profiling  | 2.0-5.0      | Low (1-10)      | Infinite        | Uniform load for consistent timing analysis               |
+| Capacity Planning  | 1.0          | Variable        | Limited         | Test resource limits with realistic constraints           |
+| SLA Validation     | 1.0          | Target rate     | SLA limit       | Production-like constraints for compliance testing        |
+
+These load patterns help evaluate different aspects of your vLLM deployment, from basic performance characteristics to resilience under challenging traffic conditions.
+
+The **Maximum Throughput** pattern (`--request-rate=inf --max-concurrency=<limit>`) is the most commonly used configuration for production benchmarking. This simulates real-world deployment architectures where:
+
+- Users send requests as fast as they can (infinite rate)
+- A load balancer or API gateway controls the maximum concurrent connections
+- The system operates at its concurrency limit, revealing true throughput capacity
+- `--burstiness` has no effect since request timing is not controlled when rate is infinite
+
+This pattern helps determine optimal concurrency settings for your production load balancer configuration.
+
+To effectively configure load patterns, especially for **Capacity Planning** and **SLA Validation** use cases, you need to understand your system's resource limits. During startup, vLLM reports KV cache configuration that directly impacts your load testing parameters:
+
+```text
+GPU KV cache size: 15,728,640 tokens
+Maximum concurrency for 8,192 tokens per request: 1920
+```
+
+Where:
+
+- GPU KV cache size: Total tokens that can be cached across all concurrent requests
+- Maximum concurrency: Theoretical maximum concurrent requests for the given `max_model_len`
+- Calculation: `max_concurrency = kv_cache_size / max_model_len`
+
+Using KV cache metrics for load pattern configuration:
+
+- For Capacity Planning: Set `--max-concurrency` to 80-90% of the reported maximum to test realistic resource constraints
+- For SLA Validation: Use the reported maximum as your SLA limit to ensure compliance testing matches production capacity
+- For Realistic Testing: Monitor memory usage when approaching theoretical limits to understand sustainable request rates
+- Request rate guidance: Use the KV cache size to estimate sustainable request rates for your specific workload and sequence lengths
+
 </details>
 
 #### 📈 Offline Throughput Benchmark
@@ -994,7 +1060,7 @@ Follow these steps to run the script:
 Example command:
 
 ```bash
-python -m vllm.benchmarks.sweep.serve \
+vllm bench sweep serve \
     --serve-cmd 'vllm serve meta-llama/Llama-2-7b-chat-hf' \
     --bench-cmd 'vllm bench serve --model meta-llama/Llama-2-7b-chat-hf --backend vllm --endpoint /v1/completions --dataset-name sharegpt --dataset-path benchmarks/ShareGPT_V3_unfiltered_cleaned_split.json' \
     --serve-params benchmarks/serve_hparams.json \
@@ -1042,7 +1108,7 @@ For example, to ensure E2E latency within different target values for 99% of req
 Example command:
 
 ```bash
-python -m vllm.benchmarks.sweep.serve_sla \
+vllm bench sweep serve_sla \
     --serve-cmd 'vllm serve meta-llama/Llama-2-7b-chat-hf' \
     --bench-cmd 'vllm bench serve --model meta-llama/Llama-2-7b-chat-hf --backend vllm --endpoint /v1/completions --dataset-name sharegpt --dataset-path benchmarks/ShareGPT_V3_unfiltered_cleaned_split.json' \
     --serve-params benchmarks/serve_hparams.json \
@@ -1071,7 +1137,7 @@ The algorithm for adjusting the SLA variable is as follows:
 Example command:
 
 ```bash
-python -m vllm.benchmarks.sweep.plot benchmarks/results/<timestamp> \
+vllm bench sweep plot benchmarks/results/<timestamp> \
     --var-x max_concurrency \
     --row-by random_input_len \
     --col-by random_output_len \
@@ -1100,7 +1166,7 @@ docker run -it --entrypoint /bin/bash -v /data/huggingface:/root/.cache/huggingf
 Then, run below command inside the docker instance.
 
 ```bash
-bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh
+bash .buildkite/performance-benchmarks/scripts/run-performance-benchmarks.sh
 ```
 
 When run, benchmark script generates results under **benchmark/results** folder, along with the benchmark_results.md and benchmark_results.json.
@@ -1118,7 +1184,7 @@ For more results visualization, check the [visualizing the results](https://gith
 
 The latest performance results are hosted on the public [vLLM Performance Dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm).
 
-More information on the performance benchmarks and their parameters can be found in [Benchmark README](https://github.com/intel-ai-tce/vllm/blob/more_cpu_models/.buildkite/nightly-benchmarks/README.md) and [performance benchmark description](../../.buildkite/nightly-benchmarks/performance-benchmarks-descriptions.md).
+More information on the performance benchmarks and their parameters can be found in [Benchmark README](https://github.com/intel-ai-tce/vllm/blob/more_cpu_models/.buildkite/nightly-benchmarks/README.md) and [performance benchmark description](../../.buildkite/performance-benchmarks/performance-benchmarks-descriptions.md).
 
 ### Continuous Benchmarking
 
@@ -1143,11 +1209,3 @@ The benchmarking currently runs on a predefined set of models configured in the
 #### Viewing Results
 
 All continuous benchmarking results are automatically published to the public [vLLM Performance Dashboard](https://hud.pytorch.org/benchmark/llms?repoName=vllm-project%2Fvllm).
-
-## Nightly Benchmarks
-
-These compare vLLM's performance against alternatives (`tgi`, `trt-llm`, and `lmdeploy`) when there are major updates of vLLM (e.g., bumping up to a new version). They are primarily intended for consumers to evaluate when to choose vLLM over other options and are triggered on every commit with both the `perf-benchmarks` and `nightly-benchmarks` labels.
-
-The latest nightly benchmark results are shared in major release blog posts such as [vLLM v0.6.0](https://blog.vllm.ai/2024/09/05/perf-update.html).
-
-More information on the nightly benchmarks and their parameters can be found [here](../../.buildkite/nightly-benchmarks/nightly-descriptions.md).

docs/deployment/docker.md

@@ -41,11 +41,11 @@ You can add any other [engine-args](../configuration/engine_args.md) you need af
     create a custom Dockerfile on top of the base image with an extra layer that installs them:
 
     ```Dockerfile
-    FROM vllm/vllm-openai:v0.9.0
+    FROM vllm/vllm-openai:v0.11.0
 
     # e.g. install the `audio` optional dependencies
     # NOTE: Make sure the version of vLLM matches the base image!
-    RUN uv pip install --system vllm[audio]==0.9.0
+    RUN uv pip install --system vllm[audio]==0.11.0
     ```
 
 !!! tip

docs/deployment/k8s.md

@@ -49,11 +49,14 @@ First, create a Kubernetes PVC and Secret for downloading and storing Hugging Fa
     metadata:
       name: hf-token-secret
     type: Opaque
-    data:
-      token: $(HF_TOKEN)
+    stringData:
+      token: "REPLACE_WITH_TOKEN"
     EOF
     ```
 
+Here, the `token` field stores your **Hugging Face access token**. For details on how to generate a token,
+see the [Hugging Face documentation](https://huggingface.co/docs/hub/en/security-tokens).
+
 Next, start the vLLM server as a Kubernetes Deployment and Service:
 
 ??? console "Config"

docs/design/io_processor_plugins.md

@@ -79,7 +79,7 @@ The `post_process*` methods take `PoolingRequestOutput` objects as input and gen
 The `validate_or_generate_params` method is used for validating with the plugin any `SamplingParameters`/`PoolingParameters` received with the user request, or to generate new ones if none are specified. The function always returns the validated/generated parameters.
 The `output_to_response` method is used only for online serving and converts the plugin output to the `IOProcessorResponse` type that is then returned by the API Server. The implementation of the `/pooling` serving endpoint is available here [vllm/entrypoints/openai/serving_pooling.py](../../vllm/entrypoints/openai/serving_pooling.py).
 
-An example implementation of a plugin that enables generating geotiff images with the PrithviGeospatialMAE model is available [here](https://github.com/IBM/terratorch/tree/main/terratorch/vllm/plugins/segmentation). Please, also refer to our online ([examples/online_serving/prithvi_geospatial_mae.py](../../examples/online_serving/prithvi_geospatial_mae.py)) and offline ([examples/offline_inference/prithvi_geospatial_mae_io_processor.py](../../examples/offline_inference/prithvi_geospatial_mae_io_processor.py)) inference examples.
+An example implementation of a plugin that enables generating geotiff images with the PrithviGeospatialMAE model is available [here](https://github.com/IBM/terratorch/tree/main/terratorch/vllm/plugins/segmentation). Please, also refer to our online ([examples/online_serving/pooling/prithvi_geospatial_mae.py](../../examples/online_serving/pooling/prithvi_geospatial_mae.py)) and offline ([examples/offline_inference/pooling/prithvi_geospatial_mae_io_processor.py](../../examples/offline_inference/pooling/prithvi_geospatial_mae_io_processor.py)) inference examples.
 
 ## Using an IO Processor plugin
 

docs/features/reasoning_outputs.md

@@ -14,11 +14,12 @@ vLLM currently supports the following reasoning models:
 | [DeepSeek-V3.1](https://huggingface.co/collections/deepseek-ai/deepseek-v31-68a491bed32bd77e7fca048f) | `deepseek_v3` | `json`, `regex` | ❌ |
 | [ERNIE-4.5-VL series](https://huggingface.co/baidu/ERNIE-4.5-VL-28B-A3B-PT) | `ernie45` | `json`, `regex` | ❌ |
 | [ERNIE-4.5-21B-A3B-Thinking](https://huggingface.co/baidu/ERNIE-4.5-21B-A3B-Thinking) | `ernie45` | `json`, `regex` | ✅ |
-| [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) | `deepseek_r1` | `json`, `regex` | ✅ |
-| [IBM Granite 3.2 language models](https://huggingface.co/collections/ibm-granite/granite-32-language-models-67b3bc8c13508f6d064cff9a) | `granite` | ❌ | ❌ |
-| [Qwen3 series](https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2e4f653967f) | `qwen3` | `json`, `regex` | ✅ |
-| [Hunyuan A13B series](https://huggingface.co/collections/tencent/hunyuan-a13b-685ec38e5b46321e3ea7c4be) | `hunyuan_a13b` | `json`, `regex` | ✅ |
 | [GLM-4.5 series](https://huggingface.co/collections/zai-org/glm-45-687c621d34bda8c9e4bf503b) | `glm45` | `json`, `regex` | ✅ |
+| [Hunyuan A13B series](https://huggingface.co/collections/tencent/hunyuan-a13b-685ec38e5b46321e3ea7c4be) | `hunyuan_a13b` | `json`, `regex` | ✅ |
+| [IBM Granite 3.2 language models](https://huggingface.co/collections/ibm-granite/granite-32-language-models-67b3bc8c13508f6d064cff9a) | `granite` | ❌ | ❌ |
+| [MiniMax-M2](https://huggingface.co/MiniMaxAI/MiniMax-M2) | `minimax_m2_append_think` | `json`, `regex` | ✅ |
+| [Qwen3 series](https://huggingface.co/collections/Qwen/qwen3-67dd247413f0e2e4f653967f) | `qwen3` | `json`, `regex` | ✅ |
+| [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) | `deepseek_r1` | `json`, `regex` | ✅ |
 
 !!! note
     IBM Granite 3.2 and DeepSeek-V3.1 reasoning is disabled by default; to enable it, you must also pass `thinking=True` in your `chat_template_kwargs`.

docs/features/tool_calling.md

@@ -321,7 +321,7 @@ Supported models:
 Flags:
 
 * For non-reasoning: `--tool-call-parser hunyuan_a13b`
-* For reasoning: `--tool-call-parser hunyuan_a13b --reasoning-parser hunyuan_a13b --enable_reasoning`
+* For reasoning: `--tool-call-parser hunyuan_a13b --reasoning-parser hunyuan_a13b`
 
 ### LongCat-Flash-Chat Models (`longcat`)
 

docs/getting_started/installation/gpu.rocm.inc.md

@@ -1,6 +1,6 @@
 # --8<-- [start:installation]
 
-vLLM supports AMD GPUs with ROCm 6.3 or above.
+vLLM supports AMD GPUs with ROCm 6.3 or above, and torch 2.8.0 and above.
 
 !!! tip
     [Docker](#set-up-using-docker) is the recommended way to use vLLM on ROCm.
@@ -28,57 +28,63 @@ Currently, there are no pre-built ROCm wheels.
 # --8<-- [end:pre-built-wheels]
 # --8<-- [start:build-wheel-from-source]
 
+!!! tip
+    - If you found that the following installation step does not work for you, please refer to [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base). Dockerfile is a form of installation steps.
+
 0. Install prerequisites (skip if you are already in an environment/docker with the following installed):
 
     - [ROCm](https://rocm.docs.amd.com/en/latest/deploy/linux/index.html)
     - [PyTorch](https://pytorch.org/)
 
-    For installing PyTorch, you can start from a fresh docker image, e.g, `rocm/pytorch:rocm6.4.3_ubuntu24.04_py3.12_pytorch_release_2.6.0`, `rocm/pytorch-nightly`. If you are using docker image, you can skip to Step 3.
+    For installing PyTorch, you can start from a fresh docker image, e.g, `rocm/pytorch:rocm7.0_ubuntu22.04_py3.10_pytorch_release_2.8.0`, `rocm/pytorch-nightly`. If you are using docker image, you can skip to Step 3.
 
     Alternatively, you can install PyTorch using PyTorch wheels. You can check PyTorch installation guide in PyTorch [Getting Started](https://pytorch.org/get-started/locally/). Example:
 
     ```bash
     # Install PyTorch
     pip uninstall torch -y
-    pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/rocm6.4
+    pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/nightly/rocm7.0
     ```
 
-1. Install [Triton for ROCm](https://github.com/triton-lang/triton)
+1. Install [Triton for ROCm](https://github.com/ROCm/triton.git)
 
-    Install ROCm's Triton (the default triton-mlir branch) following the instructions from [ROCm/triton](https://github.com/ROCm/triton/blob/triton-mlir/README.md)
+    Install ROCm's Triton following the instructions from [ROCm/triton](https://github.com/ROCm/triton.git)
 
     ```bash
     python3 -m pip install ninja cmake wheel pybind11
     pip uninstall -y triton
-    git clone https://github.com/triton-lang/triton.git
+    git clone https://github.com/ROCm/triton.git
     cd triton
-    git checkout e5be006
+    # git checkout $TRITON_BRANCH
+    git checkout f9e5bf54
     if [ ! -f setup.py ]; then cd python; fi
     python3 setup.py install
     cd ../..
     ```
 
     !!! note
-        If you see HTTP issue related to downloading packages during building triton, please try again as the HTTP error is intermittent.
+        - The validated `$TRITON_BRANCH` can be found in the [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base).
+        - If you see HTTP issue related to downloading packages during building triton, please try again as the HTTP error is intermittent.
 
-2. Optionally, if you choose to use CK flash attention, you can install [flash attention for ROCm](https://github.com/Dao-AILab/flash-attention)
+2. Optionally, if you choose to use CK flash attention, you can install [flash attention for ROCm](https://github.com/Dao-AILab/flash-attention.git)
 
-    Install ROCm's flash attention (v2.7.2) following the instructions from [ROCm/flash-attention](https://github.com/ROCm/flash-attention#amd-rocm-support)
-    Alternatively, wheels intended for vLLM use can be accessed under the releases.
+    Install ROCm's flash attention (v2.8.0) following the instructions from [ROCm/flash-attention](https://github.com/Dao-AILab/flash-attention#amd-rocm-support)
 
-    For example, for ROCm 6.3, suppose your gfx arch is `gfx90a`. To get your gfx architecture, run `rocminfo |grep gfx`.
+    For example, for ROCm 7.0, suppose your gfx arch is `gfx942`. To get your gfx architecture, run `rocminfo |grep gfx`.
 
     ```bash
     git clone https://github.com/Dao-AILab/flash-attention.git
     cd flash-attention
-    git checkout 1a7f4dfa
+    # git checkout $FA_BRANCH
+    git checkout 0e60e394
     git submodule update --init
-    GPU_ARCHS="gfx90a" python3 setup.py install
+    GPU_ARCHS="gfx942" python3 setup.py install
     cd ..
     ```
 
     !!! note
-        You might need to downgrade the "ninja" version to 1.10 as it is not used when compiling flash-attention-2 (e.g. `pip install ninja==1.10.2.4`)
+        - The validated `$FA_BRANCH` can be found in the [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base).
+
 
 3. If you choose to build AITER yourself to use a certain branch or commit, you can build AITER using the following steps:
 
@@ -92,11 +98,13 @@ Currently, there are no pre-built ROCm wheels.
     ```
 
     !!! note
-        You will need to config the `$AITER_BRANCH_OR_COMMIT` for your purpose.
+        - You will need to config the `$AITER_BRANCH_OR_COMMIT` for your purpose.
+        - The validated `$AITER_BRANCH_OR_COMMIT` can be found in the [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base).
+        
 
-4. Build vLLM. For example, vLLM on ROCM 6.3 can be built with the following steps:
+4. Build vLLM. For example, vLLM on ROCM 7.0 can be built with the following steps:
 
-    ??? console "Commands"
+    ???+ console "Commands"
 
         ```bash
         pip install --upgrade pip
@@ -109,31 +117,48 @@ Currently, there are no pre-built ROCm wheels.
             scipy \
             huggingface-hub[cli,hf_transfer] \
             setuptools_scm
-        pip install "numpy<2"
         pip install -r requirements/rocm.txt
 
-        # Build vLLM for MI210/MI250/MI300.
-        export PYTORCH_ROCM_ARCH="gfx90a;gfx942"
+        # To build for a single architecture (e.g., MI300) for faster installation (recommended):
+        export PYTORCH_ROCM_ARCH="gfx942"
+
+        # To build vLLM for multiple arch MI210/MI250/MI300, use this instead
+        # export PYTORCH_ROCM_ARCH="gfx90a;gfx942"
+
         python3 setup.py develop
         ```
 
     This may take 5-10 minutes. Currently, `pip install .` does not work for ROCm installation.
 
     !!! tip
-        - Triton flash attention is used by default. For benchmarking purposes, it is recommended to run a warm-up step before collecting perf numbers.
-        - Triton flash attention does not currently support sliding window attention. If using half precision, please use CK flash-attention for sliding window support.
-        - To use CK flash-attention or PyTorch naive attention, please use this flag `export VLLM_USE_TRITON_FLASH_ATTN=0` to turn off triton flash attention.
         - The ROCm version of PyTorch, ideally, should match the ROCm driver version.
 
 !!! tip
     - For MI300x (gfx942) users, to achieve optimal performance, please refer to [MI300x tuning guide](https://rocm.docs.amd.com/en/latest/how-to/tuning-guides/mi300x/index.html) for performance optimization and tuning tips on system and workflow level.
-      For vLLM, please refer to [vLLM performance optimization](https://rocm.docs.amd.com/en/latest/how-to/tuning-guides/mi300x/workload.html#vllm-performance-optimization).
+      For vLLM, please refer to [vLLM performance optimization](https://rocm.docs.amd.com/en/latest/how-to/rocm-for-ai/inference-optimization/vllm-optimization.html).
 
 # --8<-- [end:build-wheel-from-source]
 # --8<-- [start:pre-built-images]
 
 The [AMD Infinity hub for vLLM](https://hub.docker.com/r/rocm/vllm/tags) offers a prebuilt, optimized
 docker image designed for validating inference performance on the AMD Instinct™ MI300X accelerator.
+AMD also offers nightly prebuilt docker image from [Docker Hub](https://hub.docker.com/r/rocm/vllm-dev), which has vLLM and all its dependencies installed.
+
+???+ console "Commands"
+    ```bash
+    docker pull rocm/vllm-dev:nightly # to get the latest image
+    docker run -it --rm \
+    --network=host \
+    --group-add=video \
+    --ipc=host \
+    --cap-add=SYS_PTRACE \
+    --security-opt seccomp=unconfined \
+    --device /dev/kfd \
+    --device /dev/dri \
+    -v <path/to/your/models>:/app/models \
+    -e HF_HOME="/app/models" \
+    rocm/vllm-dev:nightly
+    ```
 
 !!! tip
     Please check [LLM inference performance validation on AMD Instinct MI300X](https://rocm.docs.amd.com/en/latest/how-to/performance-validation/mi300x/vllm-benchmark.html)
@@ -144,29 +169,29 @@ docker image designed for validating inference performance on the AMD Instinct
 
 Building the Docker image from source is the recommended way to use vLLM with ROCm.
 
-#### (Optional) Build an image with ROCm software stack
+??? info "(Optional) Build an image with ROCm software stack"
 
-Build a docker image from [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base) which setup ROCm software stack needed by the vLLM.
-**This step is optional as this rocm_base image is usually prebuilt and store at [Docker Hub](https://hub.docker.com/r/rocm/vllm-dev) under tag `rocm/vllm-dev:base` to speed up user experience.**
-If you choose to build this rocm_base image yourself, the steps are as follows.
+    Build a docker image from [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base) which setup ROCm software stack needed by the vLLM.
+    **This step is optional as this rocm_base image is usually prebuilt and store at [Docker Hub](https://hub.docker.com/r/rocm/vllm-dev) under tag `rocm/vllm-dev:base` to speed up user experience.**
+    If you choose to build this rocm_base image yourself, the steps are as follows.
 
-It is important that the user kicks off the docker build using buildkit. Either the user put DOCKER_BUILDKIT=1 as environment variable when calling docker build command, or the user needs to set up buildkit in the docker daemon configuration /etc/docker/daemon.json as follows and restart the daemon:
+    It is important that the user kicks off the docker build using buildkit. Either the user put DOCKER_BUILDKIT=1 as environment variable when calling docker build command, or the user needs to set up buildkit in the docker daemon configuration /etc/docker/daemon.json as follows and restart the daemon:
 
-```json
-{
-    "features": {
-        "buildkit": true
+    ```json
+    {
+        "features": {
+            "buildkit": true
+        }
     }
-}
-```
+    ```
 
-To build vllm on ROCm 6.3 for MI200 and MI300 series, you can use the default:
+    To build vllm on ROCm 7.0 for MI200 and MI300 series, you can use the default:
 
-```bash
-DOCKER_BUILDKIT=1 docker build \
-    -f docker/Dockerfile.rocm_base \
-    -t rocm/vllm-dev:base .
-```
+    ```bash
+    DOCKER_BUILDKIT=1 docker build \
+        -f docker/Dockerfile.rocm_base \
+        -t rocm/vllm-dev:base .
+    ```
 
 #### Build an image with vLLM
 
@@ -181,7 +206,7 @@ It is important that the user kicks off the docker build using buildkit. Either
 }
 ```
 
-[docker/Dockerfile.rocm](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm) uses ROCm 6.3 by default, but also supports ROCm 5.7, 6.0, 6.1, and 6.2, in older vLLM branches.
+[docker/Dockerfile.rocm](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm) uses ROCm 7.0 by default, but also supports ROCm 5.7, 6.0, 6.1, 6.2, 6.3, and 6.4, in older vLLM branches.
 It provides flexibility to customize the build of docker image using the following arguments:
 
 - `BASE_IMAGE`: specifies the base image used when running `docker build`. The default value `rocm/vllm-dev:base` is an image published and maintained by AMD. It is being built using [docker/Dockerfile.rocm_base](https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile.rocm_base)
@@ -189,16 +214,16 @@ It provides flexibility to customize the build of docker image using the followi
 
 Their values can be passed in when running `docker build` with `--build-arg` options.
 
-To build vllm on ROCm 6.3 for MI200 and MI300 series, you can use the default:
+To build vllm on ROCm 7.0 for MI200 and MI300 series, you can use the default:
 
-```bash
-DOCKER_BUILDKIT=1 docker build -f docker/Dockerfile.rocm -t vllm-rocm .
-```
+???+ console "Commands"
+    ```bash
+    DOCKER_BUILDKIT=1 docker build -f docker/Dockerfile.rocm -t vllm-rocm .
+    ```
 
 To run the above docker image `vllm-rocm`, use the below command:
 
-??? console "Command"
-
+???+ console "Commands"
     ```bash
     docker run -it \
     --network=host \

docs/getting_started/installation/gpu.xpu.inc.md

@@ -56,8 +56,10 @@ docker build -f docker/Dockerfile.xpu -t vllm-xpu-env --shm-size=4g .
 docker run -it \
              --rm \
              --network=host \
-             --device /dev/dri \
+             --device /dev/dri:/dev/dri \
              -v /dev/dri/by-path:/dev/dri/by-path \
+             --ipc=host \
+             --privileged \
              vllm-xpu-env
 ```
 

docs/getting_started/installation/python_env_setup.inc.md

@@ -1,4 +1,4 @@
-It's recommended to use [uv](https://docs.astral.sh/uv/), a very fast Python environment manager, to create and manage Python environments. Please follow the [documentation](https://docs.astral.sh/uv/#getting-started) to install `uv`. After installing `uv`, you can create a new Python environment using the following commands:
+On NVIDIA CUDA only, it's recommended to use [uv](https://docs.astral.sh/uv/), a very fast Python environment manager, to create and manage Python environments. Please follow the [documentation](https://docs.astral.sh/uv/#getting-started) to install `uv`. After installing `uv`, you can create a new Python environment using the following commands:
 
 ```bash
 uv venv --python 3.12 --seed

docs/mkdocs/hooks/generate_argparse.py

@@ -56,16 +56,23 @@ def auto_mock(module, attr, max_mocks=50):
     )
 
 
-latency = auto_mock("vllm.benchmarks", "latency")
-serve = auto_mock("vllm.benchmarks", "serve")
-throughput = auto_mock("vllm.benchmarks", "throughput")
+bench_latency = auto_mock("vllm.benchmarks", "latency")
+bench_serve = auto_mock("vllm.benchmarks", "serve")
+bench_sweep_plot = auto_mock("vllm.benchmarks.sweep.plot", "SweepPlotArgs")
+bench_sweep_serve = auto_mock("vllm.benchmarks.sweep.serve", "SweepServeArgs")
+bench_sweep_serve_sla = auto_mock(
+    "vllm.benchmarks.sweep.serve_sla", "SweepServeSLAArgs"
+)
+bench_throughput = auto_mock("vllm.benchmarks", "throughput")
 AsyncEngineArgs = auto_mock("vllm.engine.arg_utils", "AsyncEngineArgs")
 EngineArgs = auto_mock("vllm.engine.arg_utils", "EngineArgs")
 ChatCommand = auto_mock("vllm.entrypoints.cli.openai", "ChatCommand")
 CompleteCommand = auto_mock("vllm.entrypoints.cli.openai", "CompleteCommand")
-cli_args = auto_mock("vllm.entrypoints.openai", "cli_args")
-run_batch = auto_mock("vllm.entrypoints.openai", "run_batch")
-FlexibleArgumentParser = auto_mock("vllm.utils", "FlexibleArgumentParser")
+openai_cli_args = auto_mock("vllm.entrypoints.openai", "cli_args")
+openai_run_batch = auto_mock("vllm.entrypoints.openai", "run_batch")
+FlexibleArgumentParser = auto_mock(
+    "vllm.utils.argparse_utils", "FlexibleArgumentParser"
+)
 
 
 class MarkdownFormatter(HelpFormatter):
@@ -112,6 +119,9 @@ class MarkdownFormatter(HelpFormatter):
                 self._markdown_output.append(f"{action.help}\n\n")
 
             if (default := action.default) != SUPPRESS:
+                # Make empty string defaults visible
+                if default == "":
+                    default = '""'
                 self._markdown_output.append(f"Default: `{default}`\n\n")
 
     def format_help(self):
@@ -148,17 +158,23 @@ def on_startup(command: Literal["build", "gh-deploy", "serve"], dirty: bool):
 
     # Create parsers to document
     parsers = {
+        # Engine args
         "engine_args": create_parser(EngineArgs.add_cli_args),
         "async_engine_args": create_parser(
             AsyncEngineArgs.add_cli_args, async_args_only=True
         ),
-        "serve": create_parser(cli_args.make_arg_parser),
+        # CLI
+        "serve": create_parser(openai_cli_args.make_arg_parser),
         "chat": create_parser(ChatCommand.add_cli_args),
         "complete": create_parser(CompleteCommand.add_cli_args),
-        "bench_latency": create_parser(latency.add_cli_args),
-        "bench_throughput": create_parser(throughput.add_cli_args),
-        "bench_serve": create_parser(serve.add_cli_args),
-        "run-batch": create_parser(run_batch.make_arg_parser),
+        "run-batch": create_parser(openai_run_batch.make_arg_parser),
+        # Benchmark CLI
+        "bench_latency": create_parser(bench_latency.add_cli_args),
+        "bench_serve": create_parser(bench_serve.add_cli_args),
+        "bench_sweep_plot": create_parser(bench_sweep_plot.add_cli_args),
+        "bench_sweep_serve": create_parser(bench_sweep_serve.add_cli_args),
+        "bench_sweep_serve_sla": create_parser(bench_sweep_serve_sla.add_cli_args),
+        "bench_throughput": create_parser(bench_throughput.add_cli_args),
     }
 
     # Generate documentation for each parser

docs/models/extensions/runai_model_streamer.md

@@ -45,6 +45,15 @@ vllm serve s3://core-llm/Llama-3-8b \
 
 You can tune parameters using `--model-loader-extra-config`:
 
+You can tune `distributed` that controls whether distributed streaming should be used. This is currently only possible on CUDA and ROCM devices. This can significantly improve loading times from object storage or high-throughput network fileshares.
+You can read further about Distributed streaming [here](https://github.com/run-ai/runai-model-streamer/blob/master/docs/src/usage.md#distributed-streaming)
+
+```bash
+vllm serve /home/meta-llama/Llama-3.2-3B-Instruct \
+    --load-format runai_streamer \
+    --model-loader-extra-config '{"distributed":true}'
+```
+
 You can tune `concurrency` that controls the level of concurrency and number of OS threads reading tensors from the file to the CPU buffer.
 For reading from S3, it will be the number of client instances the host is opening to the S3 server.
 

docs/models/pooling_models.md

@@ -30,11 +30,11 @@ If `--runner pooling` has been set (manually or automatically) but the model doe
 vLLM will attempt to automatically convert the model according to the architecture names
 shown in the table below.
 
-| Architecture                                    | `--convert` | Supported pooling tasks       |
-|-------------------------------------------------|-------------|-------------------------------|
-| `*ForTextEncoding`, `*EmbeddingModel`, `*Model` | `embed`     | `encode`, `embed`             |
-| `*For*Classification`, `*ClassificationModel`   | `classify`  | `encode`, `classify`, `score` |
-| `*ForRewardModeling`, `*RewardModel`            | `reward`    | `encode`                      |
+| Architecture                                    | `--convert` | Supported pooling tasks               |
+|-------------------------------------------------|-------------|---------------------------------------|
+| `*ForTextEncoding`, `*EmbeddingModel`, `*Model` | `embed`     | `token_embed`, `embed`                |
+| `*For*Classification`, `*ClassificationModel`   | `classify`  | `token_classify`, `classify`, `score` |
+| `*ForRewardModeling`, `*RewardModel`            | `reward`    | `token_classify`                      |
 
 !!! tip
     You can explicitly set `--convert <type>` to specify how to convert the model.
@@ -45,12 +45,14 @@ Each pooling model in vLLM supports one or more of these tasks according to
 [Pooler.get_supported_tasks][vllm.model_executor.layers.pooler.Pooler.get_supported_tasks],
 enabling the corresponding APIs:
 
-| Task       | APIs                                 |
-|------------|--------------------------------------|
-| `encode`   | `LLM.reward(...)`                    |
-| `embed`    | `LLM.embed(...)`, `LLM.score(...)`\* |
-| `classify` | `LLM.classify(...)`                  |
-| `score`    | `LLM.score(...)`                     |
+| Task             | APIs                                                                          |
+|------------------|-------------------------------------------------------------------------------|
+| `embed`          | `LLM.embed(...)`, `LLM.score(...)`\*, `LLM.encode(..., pooling_task="embed")` |
+| `classify`       | `LLM.classify(...)`, `LLM.encode(..., pooling_task="classify")`               |
+| `score`          | `LLM.score(...)`                                                              |
+| `token_classify` | `LLM.reward(...)`, `LLM.encode(..., pooling_task="token_classify")`           |
+| `token_embed`    | `LLM.encode(..., pooling_task="token_embed")`                                 |
+| `plugin`         | `LLM.encode(..., pooling_task="plugin")`                                      |
 
 \* The `LLM.score(...)` API falls back to `embed` task if the model does not support `score` task.
 
@@ -144,7 +146,6 @@ A code example can be found here: [examples/offline_inference/basic/score.py](..
 ### `LLM.reward`
 
 The [reward][vllm.LLM.reward] method is available to all reward models in vLLM.
-It returns the extracted hidden states directly.
 
 ```python
 from vllm import LLM
@@ -161,15 +162,17 @@ A code example can be found here: [examples/offline_inference/basic/reward.py](.
 ### `LLM.encode`
 
 The [encode][vllm.LLM.encode] method is available to all pooling models in vLLM.
-It returns the extracted hidden states directly.
 
 !!! note
     Please use one of the more specific methods or set the task directly when using `LLM.encode`:
 
     - For embeddings, use `LLM.embed(...)` or `pooling_task="embed"`.
     - For classification logits, use `LLM.classify(...)` or `pooling_task="classify"`.
-    - For rewards, use `LLM.reward(...)` or `pooling_task="reward"`.
     - For similarity scores, use `LLM.score(...)`.  
+    - For rewards, use `LLM.reward(...)` or `pooling_task="token_classify"`.
+    - For token classification, use `pooling_task="token_classify"`.
+    - For multi-vector retrieval, use `pooling_task="token_embed"`
+    - For IO Processor Plugins , use `pooling_task="plugin"`
 
 ```python
 from vllm import LLM
@@ -185,10 +188,47 @@ print(f"Data: {data!r}")
 
 Our [OpenAI-Compatible Server](../serving/openai_compatible_server.md) provides endpoints that correspond to the offline APIs:
 
-- [Pooling API](../serving/openai_compatible_server.md#pooling-api) is similar to `LLM.encode`, being applicable to all types of pooling models.
 - [Embeddings API](../serving/openai_compatible_server.md#embeddings-api) is similar to `LLM.embed`, accepting both text and [multi-modal inputs](../features/multimodal_inputs.md) for embedding models.
 - [Classification API](../serving/openai_compatible_server.md#classification-api) is similar to `LLM.classify` and is applicable to sequence classification models.
 - [Score API](../serving/openai_compatible_server.md#score-api) is similar to `LLM.score` for cross-encoder models.
+- [Pooling API](../serving/openai_compatible_server.md#pooling-api) is similar to `LLM.encode`, being applicable to all types of pooling models.
+
+!!! note
+    Please use one of the more specific methods or set the task directly when using  [Pooling API](../serving/openai_compatible_server.md#pooling-api) api.:
+
+    - For embeddings, use [Embeddings API](../serving/openai_compatible_server.md#embeddings-api) or `"task":"embed"`.
+    - For classification logits, use [Classification API](../serving/openai_compatible_server.md#classification-api) or `task":"classify"`.
+    - For similarity scores, use [Score API](../serving/openai_compatible_server.md#score-api).  
+    - For rewards, `task":"token_classify"`.
+    - For token classification, use `task":"token_classify"`.
+    - For multi-vector retrieval, use `task":"token_embed"`
+    - For IO Processor Plugins , use `task":"plugin"`
+
+```python
+# start a supported embeddings model server with `vllm serve`, e.g.
+# vllm serve intfloat/e5-small
+import requests
+
+host = "localhost"
+port = "8000"
+model_name = "intfloat/e5-small"
+
+api_url = f"http://{host}:{port}/pooling"
+
+prompts = [
+    "Hello, my name is",
+    "The president of the United States is",
+    "The capital of France is",
+    "The future of AI is",
+]
+prompt = {"model": model_name, "input": prompts, "task": "embed"}
+
+response = requests.post(api_url, json=prompt)
+
+for output in response.json()["data"]:
+    data = output["data"]
+    print(f"Data: {data!r} (size={len(data)})")
+```
 
 ## Matryoshka Embeddings
 
@@ -265,3 +305,16 @@ Expected output:
 ```
 
 An OpenAI client example can be found here: [examples/online_serving/pooling/openai_embedding_matryoshka_fy.py](../../examples/online_serving/pooling/openai_embedding_matryoshka_fy.py)
+
+## Deprecated Features
+
+### Encode task
+
+We have split the `encode` task into two more specific token wise tasks: `token_embed` and `token_classify`:
+
+- `token_embed` is the same as embed, using normalize as activation.
+- `token_classify` is the same as classify, default using softmax as activation.
+
+### Remove softmax from PoolingParams
+
+We are going to remove `softmax` and `activation` from `PoolingParams`. Instead, you should set `use_activation`, since we actually allow `classify` and `token_classify` to use any activation function.

docs/models/supported_models.md

@@ -382,6 +382,7 @@ th {
 | `InternLM3ForCausalLM` | InternLM3 | `internlm/internlm3-8b-instruct`, etc. | ✅︎ | ✅︎ |
 | `JAISLMHeadModel` | Jais | `inceptionai/jais-13b`, `inceptionai/jais-13b-chat`, `inceptionai/jais-30b-v3`, `inceptionai/jais-30b-chat-v3`, etc. | | ✅︎ |
 | `JambaForCausalLM` | Jamba | `ai21labs/AI21-Jamba-1.5-Large`, `ai21labs/AI21-Jamba-1.5-Mini`, `ai21labs/Jamba-v0.1`, etc. | ✅︎ | ✅︎ |
+| `KimiLinearForCausalLM` | Kimi-Linear-48B-A3B-Base, Kimi-Linear-48B-A3B-Instruct | `moonshotai/Kimi-Linear-48B-A3B-Base`, `moonshotai/Kimi-Linear-48B-A3B-Instruct` | | ✅︎ |
 | `Lfm2ForCausalLM`  | LFM2  | `LiquidAI/LFM2-1.2B`, `LiquidAI/LFM2-700M`, `LiquidAI/LFM2-350M`, etc. | ✅︎ | ✅︎ |
 | `Lfm2MoeForCausalLM`  | LFM2MoE  | `LiquidAI/LFM2-8B-A1B-preview`, etc. | ✅︎ | ✅︎ |
 | `LlamaForCausalLM` | Llama 3.1, Llama 3, Llama 2, LLaMA, Yi | `meta-llama/Meta-Llama-3.1-405B-Instruct`, `meta-llama/Meta-Llama-3.1-70B`, `meta-llama/Meta-Llama-3-70B-Instruct`, `meta-llama/Llama-2-70b-hf`, `01-ai/Yi-34B`, etc. | ✅︎ | ✅︎ |
@@ -390,6 +391,7 @@ th {
 | `MiMoForCausalLM` | MiMo | `XiaomiMiMo/MiMo-7B-RL`, etc. | ✅︎ | ✅︎ |
 | `MiniCPMForCausalLM` | MiniCPM | `openbmb/MiniCPM-2B-sft-bf16`, `openbmb/MiniCPM-2B-dpo-bf16`, `openbmb/MiniCPM-S-1B-sft`, etc. | ✅︎ | ✅︎ |
 | `MiniCPM3ForCausalLM` | MiniCPM3 | `openbmb/MiniCPM3-4B`, etc. | ✅︎ | ✅︎ |
+| `MiniMaxM2ForCausalLM` | MiniMax-M2 |`MiniMaxAI/MiniMax-M2`, etc. | | ✅︎ |
 | `MistralForCausalLM` | Mistral, Mistral-Instruct | `mistralai/Mistral-7B-v0.1`, `mistralai/Mistral-7B-Instruct-v0.1`, etc. | ✅︎ | ✅︎ |
 | `MixtralForCausalLM` | Mixtral-8x7B, Mixtral-8x7B-Instruct | `mistralai/Mixtral-8x7B-v0.1`, `mistralai/Mixtral-8x7B-Instruct-v0.1`, `mistral-community/Mixtral-8x22B-v0.1`, etc. | ✅︎ | ✅︎ |
 | `MPTForCausalLM` | MPT, MPT-Instruct, MPT-Chat, MPT-StoryWriter | `mosaicml/mpt-7b`, `mosaicml/mpt-7b-storywriter`, `mosaicml/mpt-30b`, etc. | | ✅︎ |
@@ -401,6 +403,7 @@ th {
 | `OLMoEForCausalLM` | OLMoE | `allenai/OLMoE-1B-7B-0924`, `allenai/OLMoE-1B-7B-0924-Instruct`, etc. | | ✅︎ |
 | `OPTForCausalLM` | OPT, OPT-IML | `facebook/opt-66b`, `facebook/opt-iml-max-30b`, etc. | ✅︎ | ✅︎ |
 | `OrionForCausalLM` | Orion | `OrionStarAI/Orion-14B-Base`, `OrionStarAI/Orion-14B-Chat`, etc. | | ✅︎ |
+| `OuroForCausalLM` | ouro | `ByteDance/Ouro-1.4B`, `ByteDance/Ouro-2.6B`, etc. | ✅︎ | |
 | `PhiForCausalLM` | Phi | `microsoft/phi-1_5`, `microsoft/phi-2`, etc. | ✅︎ | ✅︎ |
 | `Phi3ForCausalLM` | Phi-4, Phi-3 | `microsoft/Phi-4-mini-instruct`, `microsoft/Phi-4`, `microsoft/Phi-3-mini-4k-instruct`, `microsoft/Phi-3-mini-128k-instruct`, `microsoft/Phi-3-medium-128k-instruct`, etc. | ✅︎ | ✅︎ |
 | `PhiMoEForCausalLM` | Phi-3.5-MoE | `microsoft/Phi-3.5-MoE-instruct`, etc. | ✅︎ | ✅︎ |
@@ -736,37 +739,6 @@ Some models are supported only via the [Transformers backend](#transformers). Th
 !!! note
     To use `TIGER-Lab/Mantis-8B-siglip-llama3`, you have to pass `--hf_overrides '{"architectures": ["MantisForConditionalGeneration"]}'` when running vLLM.
 
-!!! warning
-    The output quality of `AllenAI/Molmo-7B-D-0924` (especially in object localization tasks) has deteriorated in recent updates.
-
-    For the best results, we recommend using the following dependency versions (tested on A10 and L40):
-
-    ??? code "Dependency versions"
-
-        ```text
-        # Core vLLM-compatible dependencies with Molmo accuracy setup (tested on L40)
-        torch==2.5.1
-        torchvision==0.20.1
-        transformers==4.48.1
-        tokenizers==0.21.0
-        tiktoken==0.7.0
-        vllm==0.7.0
-
-        # Optional but recommended for improved performance and stability
-        triton==3.1.0
-        xformers==0.0.28.post3
-        uvloop==0.21.0
-        protobuf==5.29.3
-        openai==1.60.2
-        opencv-python-headless==4.11.0.86
-        pillow==10.4.0
-
-        # Installed FlashAttention (for float16 only)
-        flash-attn>=2.5.6  # Not used in float32, but should be documented
-        ```
-
-    **Note:** Make sure you understand the security implications of using outdated packages.
-
 !!! note
     The official `openbmb/MiniCPM-V-2` doesn't work yet, so we need to use a fork (`HwwwH/MiniCPM-V-2`) for now.
     For more details, please see: <https://github.com/vllm-project/vllm/pull/4087#issuecomment-2250397630>
@@ -805,7 +777,7 @@ The following table lists those that are tested in vLLM.
 | `CLIPModel` | CLIP | T / I | `openai/clip-vit-base-patch32`, `openai/clip-vit-large-patch14`, etc. | | |
 | `LlavaNextForConditionalGeneration`<sup>C</sup> | LLaVA-NeXT-based | T / I | `royokong/e5-v` | | ✅︎ |
 | `Phi3VForCausalLM`<sup>C</sup> | Phi-3-Vision-based | T + I | `TIGER-Lab/VLM2Vec-Full` | | ✅︎ |
-| `SiglipModel` | SigLIP | T / I | `google/siglip-base-patch16-224` | | |
+| `SiglipModel` | SigLIP, SigLIP2 | T / I | `google/siglip-base-patch16-224`, `google/siglip2-base-patch16-224` | | |
 | `*ForConditionalGeneration`<sup>C</sup>, `*ForCausalLM`<sup>C</sup>, etc. | Generative models | \* | N/A | \* | \* |
 
 <sup>C</sup> Automatically converted into an embedding model via `--convert embed`. ([details](./pooling_models.md#model-conversion))  

docs/serving/openai_compatible_server.md

@@ -638,7 +638,7 @@ Usually, the score for a sentence pair refers to the similarity between two sent
 
 You can find the documentation for cross encoder models at [sbert.net](https://www.sbert.net/docs/package_reference/cross_encoder/cross_encoder.html).
 
-Code example: [examples/online_serving/openai_cross_encoder_score.py](../../examples/online_serving/openai_cross_encoder_score.py)
+Code example: [examples/online_serving/pooling/openai_cross_encoder_score.py](../../examples/online_serving/pooling/openai_cross_encoder_score.py)
 
 #### Single inference
 
@@ -819,7 +819,7 @@ You can pass multi-modal inputs to scoring models by passing `content` including
         print("Scoring output:", response_json["data"][0]["score"])
         print("Scoring output:", response_json["data"][1]["score"])
         ```
-Full example: [examples/online_serving/openai_cross_encoder_score_for_multimodal.py](../../examples/online_serving/openai_cross_encoder_score_for_multimodal.py)
+Full example: [examples/online_serving/pooling/openai_cross_encoder_score_for_multimodal.py](../../examples/online_serving/pooling/openai_cross_encoder_score_for_multimodal.py)
 
 #### Extra parameters
 

examples/offline_inference/audio_language.py

@@ -18,7 +18,7 @@ from transformers import AutoTokenizer
 from vllm import LLM, EngineArgs, SamplingParams
 from vllm.assets.audio import AudioAsset
 from vllm.lora.request import LoRARequest
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 audio_assets = [AudioAsset("mary_had_lamb"), AudioAsset("winning_call")]
 question_per_audio_count = {

examples/offline_inference/basic/chat.py

@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from vllm import LLM, EngineArgs
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 def create_parser():

examples/offline_inference/basic/classify.py

@@ -4,7 +4,7 @@
 from argparse import Namespace
 
 from vllm import LLM, EngineArgs
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 def parse_args():

examples/offline_inference/basic/embed.py

@@ -4,7 +4,7 @@
 from argparse import Namespace
 
 from vllm import LLM, EngineArgs
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 def parse_args():

examples/offline_inference/basic/generate.py

@@ -2,7 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from vllm import LLM, EngineArgs
-from vllm.utils import FlexibleArgumentParser
+from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 
 def create_parser():