updated

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

.buildkite/run-amd-test.sh

@@ -134,9 +134,10 @@ if [[ $commands == *"--shard-id="* ]]; then
     # assign shard-id for each shard
     commands_gpu=${commands//"--shard-id= "/"--shard-id=${GPU} "}
     echo "Shard ${GPU} commands:$commands_gpu"
+    echo "Render devices: $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES"
     docker run \
-        --device /dev/kfd --device /dev/dri \
-        --network host \
+        --device /dev/kfd $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES \
+        --network=host \
         --shm-size=16gb \
         --rm \
         -e HIP_VISIBLE_DEVICES="${GPU}" \
@@ -163,9 +164,10 @@ if [[ $commands == *"--shard-id="* ]]; then
     fi
   done
 else
+  echo "Render devices: $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES"
   docker run \
-          --device /dev/kfd --device /dev/dri \
-          --network host \
+          --device /dev/kfd $BUILDKITE_AGENT_META_DATA_RENDER_DEVICES \
+          --network=host \
           --shm-size=16gb \
           --rm \
           -e HIP_VISIBLE_DEVICES=0 \

.buildkite/run-cpu-test.sh

@@ -38,6 +38,8 @@ function cpu_tests() {
     set -e
     pip install -r vllm/requirements/test.txt
     pip install -r vllm/requirements/cpu.txt
+    pytest -v -s tests/kernels/test_cache.py -m cpu_model
+    pytest -v -s tests/kernels/test_mla_decode_cpu.py -m cpu_model
     pytest -v -s tests/models/decoder_only/language -m cpu_model
     pytest -v -s tests/models/embedding/language -m cpu_model
     pytest -v -s tests/models/encoder_decoder/language -m cpu_model

.buildkite/run-gh200-test.sh

@@ -14,6 +14,7 @@ DOCKER_BUILDKIT=1 docker build . \
   -t gh200-test \
   --build-arg max_jobs=66 \
   --build-arg nvcc_threads=2 \
+  --build-arg RUN_WHEEL_CHECK=false \
   --build-arg torch_cuda_arch_list="9.0+PTX" \
   --build-arg vllm_fa_cmake_gpu_arches="90-real"
 
@@ -23,6 +24,6 @@ trap remove_docker_container EXIT
 remove_docker_container
 
 # Run the image and test offline inference
-docker run -e HF_TOKEN -v /root/.cache/huggingface:/root/.cache/huggingface --name gh200-test --gpus=all --entrypoint="" gh200-test bash -c '
+docker run -e HF_TOKEN -e VLLM_WORKER_MULTIPROC_METHOD=spawn -v /root/.cache/huggingface:/root/.cache/huggingface --name gh200-test --gpus=all --entrypoint="" gh200-test bash -c '
     python3 examples/offline_inference/basic/generate.py --model meta-llama/Llama-3.2-1B
 '

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

@@ -22,7 +22,7 @@ docker run --privileged --net host --shm-size=16G -it \
     && export VLLM_USE_V1=1 \
     && export VLLM_XLA_CHECK_RECOMPILATION=1 \
     && echo TEST_1 \
-    && python3 /workspace/vllm/tests/tpu/test_compilation.py \
+    && pytest -v -s /workspace/vllm/tests/tpu/test_compilation.py \
     && echo TEST_2 \
     && pytest -v -s /workspace/vllm/tests/v1/tpu/test_basic.py \
     && echo TEST_3 \
@@ -30,7 +30,11 @@ docker run --privileged --net host --shm-size=16G -it \
     && echo TEST_4 \
     && pytest -s -v /workspace/vllm/tests/tpu/test_quantization_accuracy.py \
     && echo TEST_5 \
-    && python3 /workspace/vllm/examples/offline_inference/tpu.py" \
+    && python3 /workspace/vllm/examples/offline_inference/tpu.py \
+    && echo TEST_6 \
+    && pytest -s -v /workspace/vllm/tests/tpu/worker/test_tpu_model_runner.py \
+    && echo TEST_7 \
+    && pytest -s -v /workspace/vllm/tests/v1/tpu/test_sampler.py" \
 
 
 # TODO: This test fails because it uses RANDOM_SEED sampling

.buildkite/test-pipeline.yaml

@@ -118,7 +118,7 @@ steps:
   - pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
   - pytest -v -s entrypoints/llm/test_generate_multiple_loras.py # it needs a clean process
   - VLLM_USE_V1=0 pytest -v -s entrypoints/llm/test_guided_generate.py # it needs a clean process
-  - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_oot_registration.py --ignore=entrypoints/openai/correctness/
+  - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_oot_registration.py  --ignore=entrypoints/openai/test_chat_with_tool_reasoning.py --ignore=entrypoints/openai/correctness/
   - pytest -v -s entrypoints/test_chat_utils.py
   - VLLM_USE_V1=0 pytest -v -s entrypoints/offline_mode # Needs to avoid interference with other tests
 
@@ -135,12 +135,14 @@ steps:
   - examples/offline_inference/rlhf.py
   - examples/offline_inference/rlhf_colocate.py
   - tests/examples/offline_inference/data_parallel.py
+  - tests/v1/test_async_llm_dp.py
   commands:
   # test with tp=2 and external_dp=2
   - VLLM_USE_V1=0 torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
   - torchrun --nproc-per-node=4 distributed/test_torchrun_example.py
   # test with internal dp
   - python3 ../examples/offline_inference/data_parallel.py
+  - TP_SIZE=2 DP_SIZE=2 pytest -v -s v1/test_async_llm_dp.py
   - pytest -v -s distributed/test_utils.py
   - pytest -v -s compile/test_basic_correctness.py
   - pytest -v -s distributed/test_pynccl.py
@@ -148,8 +150,8 @@ steps:
   # TODO: create a dedicated test section for multi-GPU example tests
   # when we have multiple distributed example tests
   - pushd ../examples/offline_inference
-  - python3 rlhf.py
-  - RAY_DEDUP_LOGS=0 python3 rlhf_colocate.py
+  - VLLM_ENABLE_V1_MULTIPROCESSING=0 python3 rlhf.py
+  - VLLM_ENABLE_V1_MULTIPROCESSING=0 RAY_DEDUP_LOGS=0 python3 rlhf_colocate.py
   - popd
 
 - label: Metrics, Tracing Test # 10min
@@ -287,7 +289,7 @@ steps:
   source_file_dependencies:
   - vllm/lora
   - tests/lora
-  command: pytest -v -s lora --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --ignore=lora/test_long_context.py --ignore=lora/test_chatglm3_tp.py --ignore=lora/test_llama_tp.py --ignore=lora/test_minicpmv_tp.py  --ignore=lora/test_transfomers_model.py
+  command: pytest -v -s lora --shard-id=$$BUILDKITE_PARALLEL_JOB --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --ignore=lora/test_chatglm3_tp.py --ignore=lora/test_llama_tp.py --ignore=lora/test_minicpmv_tp.py  --ignore=lora/test_transfomers_model.py
   parallelism: 4
 
 - label: PyTorch Fullgraph Smoke Test # 9min
@@ -514,8 +516,11 @@ steps:
   - vllm/worker/worker.py
   - vllm/worker/model_runner.py
   - entrypoints/llm/test_collective_rpc.py
+  - tests/v1/test_async_llm_dp.py
+  - vllm/v1/engine/
   commands:
-  - pytest -v -s entrypoints/llm/test_collective_rpc.py
+  - TP_SIZE=1 DP_SIZE=2 pytest -v -s v1/test_async_llm_dp.py
+  - VLLM_ENABLE_V1_MULTIPROCESSING=0 pytest -v -s entrypoints/llm/test_collective_rpc.py
   - pytest -v -s ./compile/test_basic_correctness.py
   - pytest -v -s ./compile/test_wrapper.py
   - VLLM_TEST_SAME_HOST=1 torchrun --nproc-per-node=4 distributed/test_same_node.py | grep 'Same node test passed'
@@ -592,8 +597,6 @@ steps:
     # FIXIT: find out which code initialize cuda before running the test
     # before the fix, we need to use spawn to test it
     - export VLLM_WORKER_MULTIPROC_METHOD=spawn
-    # This test runs llama 13B, so it is required to run on 4 GPUs.
-    - pytest -v -s -x lora/test_long_context.py
     # There is some Tensor Parallelism related processing logic in LoRA that
     # requires multi-GPU testing for validation.
     - pytest -v -s -x lora/test_chatglm3_tp.py

.github/mergify.yml

@@ -88,6 +88,36 @@ pull_request_rules:
       add:
         - v1
 
+- name: label-tpu
+  description: Automatically apply tpu label
+  # Keep this list in sync with `label-tpu-remove` conditions
+  conditions:
+    - or:
+      - files~=tpu.py
+      - files~=_tpu
+      - files~=tpu_
+      - files~=/tpu/
+      - files~=pallas
+  actions:
+    label:
+      add:
+        - tpu
+
+- name: label-tpu-remove
+  description: Automatically remove tpu label
+  # Keep this list in sync with `label-tpu` conditions
+  conditions:
+    - and:
+      - -files~=tpu.py
+      - -files~=_tpu
+      - -files~=tpu_
+      - -files~=/tpu/
+      - -files~=pallas
+  actions:
+    label:
+      remove:
+        - tpu
+
 - name: ping author on conflicts and add 'needs-rebase' label
   conditions:
       - conflict

.gitignore

@@ -2,7 +2,8 @@
 /vllm/_version.py
 
 # vllm-flash-attn built from source
-vllm/vllm_flash_attn/
+vllm/vllm_flash_attn/*
+!vllm/vllm_flash_attn/fa_utils.py
 
 # Byte-compiled / optimized / DLL files
 __pycache__/

CMakeLists.txt

@@ -461,6 +461,33 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
     set(FP4_ARCHS)
   endif()
 
+  #
+  # CUTLASS MoE kernels
+
+  # The MoE kernel cutlass_moe_mm requires CUDA 12.3 or later (and only works
+  # on Hopper). get_cutlass_moe_mm_data should only be compiled if it's possible
+  # to compile MoE kernels that use its output.
+  cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;" "${CUDA_ARCHS}")
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
+    set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu"
+             "csrc/quantization/cutlass_w8a8/moe/moe_data.cu")
+    set_gencode_flags_for_srcs(
+      SRCS "${SRCS}"
+      CUDA_ARCHS "${SCALED_MM_ARCHS}")
+    list(APPEND VLLM_EXT_SRC "${SRCS}")
+    list(APPEND VLLM_GPU_FLAGS "-DENABLE_CUTLASS_MOE_SM90=1")
+    message(STATUS "Building grouped_mm_c3x for archs: ${SCALED_MM_ARCHS}")
+  else()
+    if (NOT ${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
+      message(STATUS "Not building grouped_mm_c3x kernels as CUDA Compiler version is "
+                     "not >= 12.3, we recommend upgrading to CUDA 12.3 or later "
+                     "if you intend on running FP8 quantized MoE models on Hopper.")
+    else()
+      message(STATUS "Not building grouped_mm_c3x as no compatible archs found "
+                     "in CUDA target architectures")
+    endif()
+  endif()
+
   #
   # Machete kernels
 

Dockerfile

@@ -14,17 +14,22 @@ ARG PYTHON_VERSION=3.12
 ARG TARGETPLATFORM
 ENV DEBIAN_FRONTEND=noninteractive
 
-# Install minimal dependencies and uv
-RUN apt-get update -y \
-    && apt-get install -y ccache git curl wget sudo \
-    && curl -LsSf https://astral.sh/uv/install.sh | sh
-
-# Add uv to PATH
-ENV PATH="/root/.local/bin:$PATH"
-# Create venv with specified Python and activate by placing at the front of path
-ENV VIRTUAL_ENV="/opt/venv"
-RUN uv venv --python ${PYTHON_VERSION} --seed ${VIRTUAL_ENV}
-ENV PATH="$VIRTUAL_ENV/bin:$PATH"
+# Install Python and other dependencies
+RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \
+    && echo 'tzdata tzdata/Zones/America select Los_Angeles' | debconf-set-selections \
+    && apt-get update -y \
+    && apt-get install -y ccache software-properties-common git curl sudo \
+    && add-apt-repository ppa:deadsnakes/ppa \
+    && apt-get update -y \
+    && apt-get install -y python${PYTHON_VERSION} python${PYTHON_VERSION}-dev python${PYTHON_VERSION}-venv \
+    && update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1 \
+    && update-alternatives --set python3 /usr/bin/python${PYTHON_VERSION} \
+    && ln -sf /usr/bin/python${PYTHON_VERSION}-config /usr/bin/python3-config \
+    && curl -sS https://bootstrap.pypa.io/get-pip.py | python${PYTHON_VERSION} \
+    && python3 --version && python3 -m pip --version
+# Install uv for faster pip installs
+RUN --mount=type=cache,target=/root/.cache/uv \
+    python3 -m pip install uv
 
 # This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
 # Reference: https://github.com/astral-sh/uv/pull/1694
@@ -46,19 +51,22 @@ RUN ldconfig /usr/local/cuda-$(echo $CUDA_VERSION | cut -d. -f1,2)/compat/
 
 WORKDIR /workspace
 
+# install build and runtime dependencies
+
 # arm64 (GH200) build follows the practice of "use existing pytorch" build,
 # we need to install torch and torchvision from the nightly builds first,
 # pytorch will not appear as a vLLM dependency in all of the following steps
 # after this step
 RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "$TARGETPLATFORM" = "linux/arm64" ]; then \
-        uv pip install --index-url https://download.pytorch.org/whl/nightly/cu126 "torch==2.7.0.dev20250121+cu126" "torchvision==0.22.0.dev20250121";  \
+        uv pip install --system --index-url https://download.pytorch.org/whl/nightly/cu128 "torch==2.8.0.dev20250318+cu128" "torchvision==0.22.0.dev20250319";  \
+        uv pip install --system --index-url https://download.pytorch.org/whl/nightly/cu128 --pre pytorch_triton==3.3.0+gitab727c40; \
     fi
 
 COPY requirements/common.txt requirements/common.txt
 COPY requirements/cuda.txt requirements/cuda.txt
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/cuda.txt
+    uv pip install --system -r requirements/cuda.txt
 
 # cuda arch list used by torch
 # can be useful for both `dev` and `test`
@@ -83,7 +91,7 @@ COPY requirements/build.txt requirements/build.txt
 ENV UV_HTTP_TIMEOUT=500
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/build.txt
+    uv pip install --system -r requirements/build.txt
 
 COPY . .
 ARG GIT_REPO_CHECK=0
@@ -155,7 +163,7 @@ COPY requirements/lint.txt requirements/lint.txt
 COPY requirements/test.txt requirements/test.txt
 COPY requirements/dev.txt requirements/dev.txt
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/dev.txt
+    uv pip install --system -r requirements/dev.txt
 #################### DEV IMAGE ####################
 
 #################### vLLM installation IMAGE ####################
@@ -171,18 +179,23 @@ ARG TARGETPLATFORM
 RUN PYTHON_VERSION_STR=$(echo ${PYTHON_VERSION} | sed 's/\.//g') && \
     echo "export PYTHON_VERSION_STR=${PYTHON_VERSION_STR}" >> /etc/environment
 
-# Install minimal dependencies and uv
-RUN apt-get update -y \
-    && apt-get install -y ccache git curl wget sudo vim \
-    && apt-get install -y ffmpeg libsm6 libxext6 libgl1 libibverbs-dev \
-    && curl -LsSf https://astral.sh/uv/install.sh | sh
-
-# Add uv to PATH
-ENV PATH="/root/.local/bin:$PATH"
-# Create venv with specified Python and activate by placing at the front of path
-ENV VIRTUAL_ENV="/opt/venv"
-RUN uv venv --python ${PYTHON_VERSION} --seed ${VIRTUAL_ENV}
-ENV PATH="$VIRTUAL_ENV/bin:$PATH"
+# Install Python and other dependencies
+RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \
+    && echo 'tzdata tzdata/Zones/America select Los_Angeles' | debconf-set-selections \
+    && apt-get update -y \
+    && apt-get install -y ccache software-properties-common git curl wget sudo vim python3-pip \
+    && apt-get install -y ffmpeg libsm6 libxext6 libgl1 \
+    && add-apt-repository ppa:deadsnakes/ppa \
+    && apt-get update -y \
+    && apt-get install -y python${PYTHON_VERSION} python${PYTHON_VERSION}-dev python${PYTHON_VERSION}-venv libibverbs-dev \
+    && update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1 \
+    && update-alternatives --set python3 /usr/bin/python${PYTHON_VERSION} \
+    && ln -sf /usr/bin/python${PYTHON_VERSION}-config /usr/bin/python3-config \
+    && curl -sS https://bootstrap.pypa.io/get-pip.py | python${PYTHON_VERSION} \
+    && python3 --version && python3 -m pip --version
+# Install uv for faster pip installs
+RUN --mount=type=cache,target=/root/.cache/uv \
+    python3 -m pip install uv
 
 # This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
 # Reference: https://github.com/astral-sh/uv/pull/1694
@@ -200,13 +213,14 @@ RUN ldconfig /usr/local/cuda-$(echo $CUDA_VERSION | cut -d. -f1,2)/compat/
 # after this step
 RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "$TARGETPLATFORM" = "linux/arm64" ]; then \
-        uv pip install --index-url https://download.pytorch.org/whl/nightly/cu124 "torch==2.6.0.dev20241210+cu124" "torchvision==0.22.0.dev20241215";  \
+        uv pip install --system --index-url https://download.pytorch.org/whl/nightly/cu128 "torch==2.8.0.dev20250318+cu128" "torchvision==0.22.0.dev20250319";  \
+        uv pip install --system --index-url https://download.pytorch.org/whl/nightly/cu128 --pre pytorch_triton==3.3.0+gitab727c40; \
     fi
 
 # Install vllm wheel first, so that torch etc will be installed.
 RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist \
     --mount=type=cache,target=/root/.cache/uv \
-    uv pip install dist/*.whl --verbose
+    uv pip install --system dist/*.whl --verbose
 
 # If we need to build FlashInfer wheel before its release:
 # $ export FLASHINFER_ENABLE_AOT=1
@@ -221,8 +235,9 @@ RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist
 # $ # upload the wheel to a public location, e.g. https://wheels.vllm.ai/flashinfer/524304395bd1d8cd7d07db083859523fcaa246a4/flashinfer_python-0.2.1.post1+cu124torch2.5-cp38-abi3-linux_x86_64.whl
 
 RUN --mount=type=cache,target=/root/.cache/uv \
+. /etc/environment && \
 if [ "$TARGETPLATFORM" != "linux/arm64" ]; then \
-    uv pip install https://github.com/flashinfer-ai/flashinfer/releases/download/v0.2.1.post2/flashinfer_python-0.2.1.post2+cu124torch2.6-cp38-abi3-linux_x86_64.whl ; \
+    uv pip install --system https://github.com/flashinfer-ai/flashinfer/releases/download/v0.2.1.post2/flashinfer_python-0.2.1.post2+cu124torch2.6-cp38-abi3-linux_x86_64.whl ; \
 fi
 COPY examples examples
 
@@ -232,7 +247,7 @@ COPY examples examples
 # TODO: Remove this once FlashInfer AOT wheel is fixed
 COPY requirements/build.txt requirements/build.txt
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/build.txt
+    uv pip install --system -r requirements/build.txt
 
 #################### vLLM installation IMAGE ####################
 
@@ -249,15 +264,15 @@ ENV UV_HTTP_TIMEOUT=500
 
 # install development dependencies (for testing)
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/dev.txt
+    uv pip install --system -r requirements/dev.txt
 
 # install development dependencies (for testing)
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -e tests/vllm_test_utils
+    uv pip install --system -e tests/vllm_test_utils
 
 # enable fast downloads from hf (for testing)
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install hf_transfer
+    uv pip install --system hf_transfer
 ENV HF_HUB_ENABLE_HF_TRANSFER 1
 
 # Copy in the v1 package for testing (it isn't distributed yet)
@@ -282,9 +297,9 @@ ENV UV_HTTP_TIMEOUT=500
 # install additional dependencies for openai api server
 RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "$TARGETPLATFORM" = "linux/arm64" ]; then \
-        uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.42.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
+        uv pip install --system accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.42.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
     else \
-        uv pip install accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.45.0' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
+        uv pip install --system accelerate hf_transfer 'modelscope!=1.15.0' 'bitsandbytes>=0.45.3' 'timm==0.9.10' boto3 runai-model-streamer runai-model-streamer[s3]; \
     fi
 
 ENV VLLM_USAGE_SOURCE production-docker-image

Dockerfile.ppc64le

@@ -1,37 +1,267 @@
-FROM mambaorg/micromamba
-ARG MAMBA_DOCKERFILE_ACTIVATE=1
-USER root
+ARG BASE_UBI_IMAGE_TAG=9.5-1741850109
 
-ENV PATH="/usr/local/cargo/bin:$PATH:/opt/conda/bin/"
+###############################################################
+# base stage with basic dependencies
+###############################################################
 
-RUN apt-get update -y && apt-get install -y git wget kmod curl vim libnuma-dev libsndfile-dev libprotobuf-dev build-essential ffmpeg libsm6 libxext6 libgl1 libssl-dev 
+FROM registry.access.redhat.com/ubi9/ubi-minimal:${BASE_UBI_IMAGE_TAG} AS base-builder
 
-# Some packages in requirements/cpu are installed here
-# IBM provides optimized packages for ppc64le processors in the open-ce project for mamba
-# Currently these may not be available for venv or pip directly
-RUN micromamba install -y -n base -c https://ftp.osuosl.org/pub/open-ce/1.11.0-p10/ -c defaults python=3.10 rust && micromamba clean --all --yes
+ARG PYTHON_VERSION=3.12
+ARG OPENBLAS_VERSION=0.3.29
+
+# Set Environment Variables for venv, cargo & openblas
+ENV VIRTUAL_ENV=/opt/vllm
+ENV PATH=${VIRTUAL_ENV}/bin:/root/.cargo/bin:$PATH
+ENV PKG_CONFIG_PATH=/usr/local/lib/pkgconfig/
+ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib64:/usr/local/lib:/usr/lib64:/usr/lib
+ENV UV_LINK_MODE=copy
+
+# install gcc-13, python, rust, openblas
+# Note: A symlink for libatomic.so is created for gcc-13 (linker fails to find libatomic otherwise - reqd. for sentencepiece)
+# Note: A dummy file 'control' is created in /tmp/ to artificially create dependencies between stages when building stages in parallel
+#       when `--jobs=<N>` is passed with podman build command
+RUN microdnf install -y openssl-devel dnf \
+    && dnf install -y https://mirror.stream.centos.org/9-stream/BaseOS/`arch`/os/Packages/centos-gpg-keys-9.0-24.el9.noarch.rpm \
+        https://mirror.stream.centos.org/9-stream/BaseOS/`arch`/os/Packages/centos-stream-repos-9.0-24.el9.noarch.rpm \
+        https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm \
+    && dnf config-manager --add-repo https://mirror.stream.centos.org/9-stream/BaseOS/`arch`/os \
+    && dnf config-manager --add-repo https://mirror.stream.centos.org/9-stream/AppStream/`arch`/os \
+    && dnf config-manager --set-enabled crb \
+    && dnf install -y \
+       git tar gcc-toolset-13 automake libtool numactl-devel lapack-devel \
+       pkgconfig xsimd zeromq-devel kmod findutils protobuf* \
+       libtiff-devel libjpeg-devel openjpeg2-devel zlib-devel \
+       freetype-devel lcms2-devel libwebp-devel tcl-devel tk-devel \
+       harfbuzz-devel fribidi-devel libraqm-devel libimagequant-devel libxcb-devel \
+       python${PYTHON_VERSION}-devel python${PYTHON_VERSION}-pip \
+    && dnf clean all \
+    && ln -sf /usr/lib64/libatomic.so.1 /usr/lib64/libatomic.so \
+    && python${PYTHON_VERSION} -m venv ${VIRTUAL_ENV} \
+    && python -m pip install -U pip uv \
+    && uv pip install wheel build "setuptools<70" setuptools_scm setuptools_rust meson-python cmake ninja cython scikit_build_core scikit_build \
+    && curl -sL https://ftp2.osuosl.org/pub/ppc64el/openblas/latest/Openblas_${OPENBLAS_VERSION}_ppc64le.tar.gz | tar xvf - -C /usr/local \
+    && curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y \
+    && cd /tmp && touch control
+
+###############################################################
+# Stage to build torch family
+###############################################################
+
+FROM base-builder AS torch-builder
+
+ARG MAX_JOBS
+ARG TORCH_VERSION=2.6.0
+ARG _GLIBCXX_USE_CXX11_ABI=1
+RUN --mount=type=cache,target=/root/.cache/uv \
+    source /opt/rh/gcc-toolset-13/enable &&  \
+    git clone --recursive https://github.com/pytorch/pytorch.git -b v${TORCH_VERSION} && \
+    cd pytorch && \
+    uv pip install -r requirements.txt && \
+    python setup.py develop && \
+    rm -f dist/torch*+git*whl && \
+    MAX_JOBS=${MAX_JOBS:-$(nproc)} \
+    PYTORCH_BUILD_VERSION=${TORCH_VERSION} PYTORCH_BUILD_NUMBER=1 uv build --wheel --out-dir /torchwheels/
+
+ARG TORCHVISION_VERSION=0.21.0
+ARG TORCHVISION_USE_NVJPEG=0
+ARG TORCHVISION_USE_FFMPEG=0
+RUN --mount=type=cache,target=/root/.cache/uv \
+    source /opt/rh/gcc-toolset-13/enable && \
+    git clone --recursive https://github.com/pytorch/vision.git -b v${TORCHVISION_VERSION} && \
+    cd vision && \
+    MAX_JOBS=${MAX_JOBS:-$(nproc)} \
+    BUILD_VERSION=${TORCHVISION_VERSION} \
+    uv build --wheel --out-dir /torchwheels/ --no-build-isolation
+
+ARG TORCHAUDIO_VERSION=2.6.0
+ARG BUILD_SOX=1
+ARG BUILD_KALDI=1
+ARG BUILD_RNNT=1
+ARG USE_FFMPEG=0
+ARG USE_ROCM=0
+ARG USE_CUDA=0
+ARG TORCHAUDIO_TEST_ALLOW_SKIP_IF_NO_FFMPEG=1
+RUN --mount=type=cache,target=/root/.cache/uv \
+    source /opt/rh/gcc-toolset-13/enable && \
+    git clone --recursive https://github.com/pytorch/audio.git -b v${TORCHAUDIO_VERSION} && \
+    cd audio && \
+    MAX_JOBS=${MAX_JOBS:-$(nproc)} \
+    BUILD_VERSION=${TORCHAUDIO_VERSION} \
+    uv build --wheel --out-dir /torchwheels/ --no-build-isolation
+
+###############################################################
+# Stage to build pyarrow
+###############################################################
+
+FROM base-builder AS arrow-builder
+
+ARG MAX_JOBS
+ARG PYARROW_PARALLEL
+ARG PYARROW_VERSION=19.0.1
+RUN --mount=type=cache,target=/root/.cache/uv \
+    source /opt/rh/gcc-toolset-13/enable && \
+    git clone --recursive https://github.com/apache/arrow.git -b apache-arrow-${PYARROW_VERSION} && \
+    cd arrow/cpp && \
+    mkdir build && cd build && \
+    cmake -DCMAKE_BUILD_TYPE=release \
+        -DCMAKE_INSTALL_PREFIX=/usr/local \
+        -DARROW_PYTHON=ON \
+        -DARROW_BUILD_TESTS=OFF \
+        -DARROW_JEMALLOC=ON \
+        -DARROW_BUILD_STATIC="OFF" \
+        -DARROW_PARQUET=ON \
+        .. && \
+    make install -j ${MAX_JOBS:-$(nproc)} && \
+    cd ../../python/ && \
+    uv pip install -v -r requirements-wheel-build.txt && \
+    PYARROW_PARALLEL=${PYARROW_PARALLEL:-$(nproc)} \
+    python setup.py build_ext \
+    --build-type=release --bundle-arrow-cpp \
+    bdist_wheel --dist-dir /arrowwheels/
+
+###############################################################
+# Stage to build opencv
+###############################################################
+
+FROM base-builder AS cv-builder
+
+ARG MAX_JOBS
+ARG OPENCV_VERSION=84
+ARG ENABLE_HEADLESS=1
+RUN --mount=type=cache,target=/root/.cache/uv \
+    source /opt/rh/gcc-toolset-13/enable && \
+    git clone --recursive https://github.com/opencv/opencv-python.git -b ${OPENCV_VERSION} && \
+    cd opencv-python && \
+    sed -i 's/"setuptools==59.2.0",/"setuptools<70.0",/g' pyproject.toml && \
+    python -m build --wheel --installer=uv --outdir /opencvwheels/
+
+###############################################################
+# Stage to build vllm - this stage builds and installs
+# vllm, tensorizer and vllm-tgis-adapter and builds uv cache
+# for transitive dependencies - eg. grpcio
+###############################################################
+
+FROM base-builder AS vllmcache-builder
+
+COPY --from=torch-builder /tmp/control /dev/null
+COPY --from=arrow-builder /tmp/control /dev/null
+COPY --from=cv-builder /tmp/control /dev/null
+
+ARG VLLM_TARGET_DEVICE=cpu
+
+# this step installs vllm and populates uv cache
+# with all the transitive dependencies
+RUN --mount=type=cache,target=/root/.cache/uv \
+    --mount=type=bind,from=torch-builder,source=/torchwheels/,target=/torchwheels/,ro \
+    --mount=type=bind,from=arrow-builder,source=/arrowwheels/,target=/arrowwheels/,ro \
+    --mount=type=bind,from=cv-builder,source=/opencvwheels/,target=/opencvwheels/,ro \
+    --mount=type=bind,src=.,dst=/src/,rw \
+    source /opt/rh/gcc-toolset-13/enable && \
+    uv pip install /opencvwheels/*.whl /arrowwheels/*.whl /torchwheels/*.whl && \
+    sed -i -e 's/.*torch.*//g' /src/pyproject.toml /src/requirements/*.txt && \
+    uv pip install pandas pythran pybind11 && \
+    # sentencepiece.pc is in some pkgconfig inside uv cache
+    export PKG_CONFIG_PATH=$(find / -type d -name "pkgconfig" 2>/dev/null | tr '\n' ':') && \
+    uv pip install -r /src/requirements/common.txt -r /src/requirements/cpu.txt -r /src/requirements/build.txt --no-build-isolation && \
+    cd /src/ && \
+    uv build --wheel --out-dir /vllmwheel/ --no-build-isolation && \
+    uv pip install /vllmwheel/*.whl
+
+
+###############################################################
+# Stage to build numactl
+###############################################################
+
+FROM base-builder AS numa-builder
+
+# Note: Building numactl with gcc-11. Compiling with gcc-13 in this builder stage will
+# trigger recompilation with gcc-11 (and require libtool) in the final stage where we do not have gcc-13
+ARG MAX_JOBS
+ARG NUMACTL_VERSION=2.0.19
+RUN git clone --recursive https://github.com/numactl/numactl.git -b v${NUMACTL_VERSION} \
+    && cd numactl \
+    && autoreconf -i && ./configure \
+    && make -j ${MAX_JOBS:-$(nproc)}
+
+###############################################################
+# Stage to build lapack
+###############################################################
+
+FROM base-builder AS lapack-builder
+
+ARG MAX_JOBS
+ARG LAPACK_VERSION=3.12.1
+RUN git clone --recursive https://github.com/Reference-LAPACK/lapack.git -b v${LAPACK_VERSION} \
+    && cd lapack && source /opt/rh/gcc-toolset-13/enable \
+    && cmake -B build -S . \
+    && cmake --build build -j ${MAX_JOBS:-$(nproc)}
+
+
+###############################################################
+#                   FINAL VLLM IMAGE STAGE                    #
+###############################################################
+
+FROM registry.access.redhat.com/ubi9/ubi-minimal:${BASE_UBI_IMAGE_TAG} AS vllm-openai
+
+ARG PYTHON_VERSION=3.12
+ARG OPENBLAS_VERSION=0.3.29
+
+# Set Environment Variables for venv & openblas
+ENV VIRTUAL_ENV=/opt/vllm
+ENV PATH=${VIRTUAL_ENV}/bin:$PATH
+ENV PKG_CONFIG_PATH=/usr/local/lib/pkgconfig/
+ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib64:/usr/local/lib:/usr/lib64:/usr/lib
+ENV UV_LINK_MODE=copy
+
+# create artificial dependencies between stages for independent stages to build in parallel
+COPY --from=torch-builder /tmp/control /dev/null
+COPY --from=arrow-builder /tmp/control /dev/null
+COPY --from=cv-builder /tmp/control /dev/null
+COPY --from=vllmcache-builder /tmp/control /dev/null
+COPY --from=numa-builder /tmp/control /dev/null
+COPY --from=lapack-builder /tmp/control /dev/null
+
+# install gcc-11, python, openblas, numactl, lapack
+RUN --mount=type=cache,target=/root/.cache/uv \
+    --mount=type=bind,from=numa-builder,source=/numactl/,target=/numactl/,rw \
+    --mount=type=bind,from=lapack-builder,source=/lapack/,target=/lapack/,rw \
+    rpm -ivh https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm && \
+    microdnf install --nodocs -y \
+    tar findutils openssl \
+    pkgconfig xsimd g++ gcc-fortran libsndfile \
+    libtiff libjpeg openjpeg2 zlib zeromq \
+    freetype lcms2 libwebp tcl tk utf8proc \
+    harfbuzz fribidi libraqm libimagequant libxcb \
+    python${PYTHON_VERSION}-devel python${PYTHON_VERSION}-pip \
+    && microdnf clean all \
+    && python${PYTHON_VERSION} -m venv ${VIRTUAL_ENV} \
+    && python -m pip install -U pip uv --no-cache \
+    && curl -sL https://ftp2.osuosl.org/pub/ppc64el/openblas/latest/Openblas_${OPENBLAS_VERSION}_ppc64le.tar.gz | tar xvf - -C /usr/local \
+    && make -C /numactl install \
+    && uv pip install cmake \
+    && cmake --install /lapack/build \
+    && uv pip uninstall cmake
+
+# consume previously built wheels (including vllm)
+RUN --mount=type=cache,target=/root/.cache/uv \
+    --mount=type=bind,from=torch-builder,source=/torchwheels/,target=/torchwheels/,ro \
+    --mount=type=bind,from=arrow-builder,source=/arrowwheels/,target=/arrowwheels/,ro \
+    --mount=type=bind,from=cv-builder,source=/opencvwheels/,target=/opencvwheels/,ro \
+    --mount=type=bind,from=vllmcache-builder,source=/vllmwheel/,target=/vllmwheel/,ro \
+    HOME=/root uv pip install /opencvwheels/*.whl /arrowwheels/*.whl /torchwheels/*.whl /vllmwheel/*.whl
 
 COPY ./ /workspace/vllm
-
 WORKDIR /workspace/vllm
 ARG GIT_REPO_CHECK=0
 RUN --mount=type=bind,source=.git,target=.git \
     if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh; fi
 
-RUN --mount=type=cache,target=/root/.cache/pip  \
-    RUSTFLAGS='-L /opt/conda/lib' pip install -v --prefer-binary --extra-index-url https://repo.fury.io/mgiessing \
-        'cmake>=3.26' ninja packaging 'setuptools-scm>=8' wheel jinja2 \
-        -r requirements/cpu.txt \
-        xformers uvloop==0.20.0
-
-RUN --mount=type=bind,source=.git,target=.git \
-    VLLM_TARGET_DEVICE=cpu python3 setup.py install
-
 # install development dependencies (for testing)
-RUN python3 -m pip install -e tests/vllm_test_utils
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv pip install -e tests/vllm_test_utils
 
 WORKDIR /workspace/
 
 RUN ln -s /workspace/vllm/tests && ln -s /workspace/vllm/examples && ln -s /workspace/vllm/benchmarks
 
-ENTRYPOINT ["/opt/conda/bin/python3", "-m", "vllm.entrypoints.openai.api_server"]
+ENTRYPOINT ["python", "-m", "vllm.entrypoints.openai.api_server"]

Dockerfile.rocm

@@ -12,7 +12,8 @@ ENV PYTORCH_ROCM_ARCH=${ARG_PYTORCH_ROCM_ARCH:-${PYTORCH_ROCM_ARCH}}
 
 # Install some basic utilities
 RUN apt-get update -q -y && apt-get install -q -y \
-    sqlite3 libsqlite3-dev libfmt-dev libmsgpack-dev libsuitesparse-dev
+    sqlite3 libsqlite3-dev libfmt-dev libmsgpack-dev libsuitesparse-dev \
+    apt-transport-https ca-certificates wget curl
 # Remove sccache    
 RUN python3 -m pip install --upgrade pip && pip install setuptools_scm
 RUN apt-get purge -y sccache; python3 -m pip uninstall -y sccache; rm -f "$(which sccache)"
@@ -40,7 +41,7 @@ ARG USE_CYTHON
 RUN cd vllm \
     && python3 -m pip install -r requirements/rocm.txt \
     && python3 setup.py clean --all  \
-    && if [ ${USE_CYTHON} -eq "1" ]; then python3 setup_cython.py build_ext --inplace; fi \
+    && if [ ${USE_CYTHON} -eq "1" ]; then python3 tests/build_cython.py build_ext --inplace; fi \
     && python3 setup.py bdist_wheel --dist-dir=dist
 FROM scratch AS export_vllm
 ARG COMMON_WORKDIR

README.md

@@ -28,10 +28,27 @@ Easy, fast, and cheap LLM serving for everyone
 - [2025/02] We hosted [the ninth vLLM meetup](https://lu.ma/h7g3kuj9) with Meta! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1jzC_PZVXrVNSFVCW-V4cFXb6pn7zZ2CyP_Flwo05aqg/edit?usp=sharing) and AMD [here](https://drive.google.com/file/d/1Zk5qEJIkTmlQ2eQcXQZlljAx3m9s7nwn/view?usp=sharing). The slides from Meta will not be posted.
 - [2025/01] We are excited to announce the alpha release of vLLM V1: A major architectural upgrade with 1.7x speedup! Clean code, optimized execution loop, zero-overhead prefix caching, enhanced multimodal support, and more. Please check out our blog post [here](https://blog.vllm.ai/2025/01/27/v1-alpha-release.html).
 - [2025/01] We hosted [the eighth vLLM meetup](https://lu.ma/zep56hui) with Google Cloud! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1epVkt4Zu8Jz_S5OhEHPc798emsYh2BwYfRuDDVEF7u4/edit?usp=sharing), and Google Cloud team [here](https://drive.google.com/file/d/1h24pHewANyRL11xy5dXUbvRC9F9Kkjix/view?usp=sharing).
-- [2024/12] vLLM joins [PyTorch ecosystem](https://pytorch.org/blog/vllm-joins-pytorch)! Easy, Fast, and Cheap LLM Serving for Everyone!
+- [2024/12] vLLM joins [pytorch ecosystem](https://pytorch.org/blog/vllm-joins-pytorch)! Easy, Fast, and Cheap LLM Serving for Everyone!
+
+<details>
+<summary>Previous News</summary>
+
+- [2024/11] We hosted [the seventh vLLM meetup](https://lu.ma/h0qvrajz) with Snowflake! Please find the meetup slides from vLLM team [here](https://docs.google.com/presentation/d/1e3CxQBV3JsfGp30SwyvS3eM_tW-ghOhJ9PAJGK6KR54/edit?usp=sharing), and Snowflake team [here](https://docs.google.com/presentation/d/1qF3RkDAbOULwz9WK5TOltt2fE9t6uIc_hVNLFAaQX6A/edit?usp=sharing).
+- [2024/10] We have just created a developer slack ([slack.vllm.ai](https://slack.vllm.ai)) focusing on coordinating contributions and discussing features. Please feel free to join us there!
+- [2024/10] Ray Summit 2024 held a special track for vLLM! Please find the opening talk slides from the vLLM team [here](https://docs.google.com/presentation/d/1B_KQxpHBTRa_mDF-tR6i8rWdOU5QoTZNcEg2MKZxEHM/edit?usp=sharing). Learn more from the [talks](https://www.youtube.com/playlist?list=PLzTswPQNepXl6AQwifuwUImLPFRVpksjR) from other vLLM contributors and users!
+- [2024/09] We hosted [the sixth vLLM meetup](https://lu.ma/87q3nvnh) with NVIDIA! Please find the meetup slides [here](https://docs.google.com/presentation/d/1wrLGwytQfaOTd5wCGSPNhoaW3nq0E-9wqyP7ny93xRs/edit?usp=sharing).
+- [2024/07] We hosted [the fifth vLLM meetup](https://lu.ma/lp0gyjqr) with AWS! Please find the meetup slides [here](https://docs.google.com/presentation/d/1RgUD8aCfcHocghoP3zmXzck9vX3RCI9yfUAB2Bbcl4Y/edit?usp=sharing).
+- [2024/07] In partnership with Meta, vLLM officially supports Llama 3.1 with FP8 quantization and pipeline parallelism! Please check out our blog post [here](https://blog.vllm.ai/2024/07/23/llama31.html).
+- [2024/06] We hosted [the fourth vLLM meetup](https://lu.ma/agivllm) with Cloudflare and BentoML! Please find the meetup slides [here](https://docs.google.com/presentation/d/1iJ8o7V2bQEi0BFEljLTwc5G1S10_Rhv3beed5oB0NJ4/edit?usp=sharing).
+- [2024/04] We hosted [the third vLLM meetup](https://robloxandvllmmeetup2024.splashthat.com/) with Roblox! Please find the meetup slides [here](https://docs.google.com/presentation/d/1A--47JAK4BJ39t954HyTkvtfwn0fkqtsL8NGFuslReM/edit?usp=sharing).
+- [2024/01] We hosted [the second vLLM meetup](https://lu.ma/ygxbpzhl) with IBM! Please find the meetup slides [here](https://docs.google.com/presentation/d/12mI2sKABnUw5RBWXDYY-HtHth4iMSNcEoQ10jDQbxgA/edit?usp=sharing).
+- [2023/10] We hosted [the first vLLM meetup](https://lu.ma/first-vllm-meetup) with a16z! Please find the meetup slides [here](https://docs.google.com/presentation/d/1QL-XPFXiFpDBh86DbEegFXBXFXjix4v032GhShbKf3s/edit?usp=sharing).
+- [2023/08] We would like to express our sincere gratitude to [Andreessen Horowitz](https://a16z.com/2023/08/30/supporting-the-open-source-ai-community/) (a16z) for providing a generous grant to support the open-source development and research of vLLM.
+- [2023/06] We officially released vLLM! FastChat-vLLM integration has powered [LMSYS Vicuna and Chatbot Arena](https://chat.lmsys.org) since mid-April. Check out our [blog post](https://vllm.ai).
+
+</details>
 
 ---
-
 ## About
 
 vLLM is a fast and easy-to-use library for LLM inference and serving.

benchmarks/kernels/benchmark_grouped_gemm_cutlass.py

@@ -0,0 +1,340 @@
+# SPDX-License-Identifier: Apache-2.0
+
+import torch
+import torch.utils.benchmark as benchmark
+from benchmark_shapes import WEIGHT_SHAPES_MOE
+
+from vllm import _custom_ops as ops
+from vllm.config import ParallelConfig, VllmConfig, set_current_vllm_config
+from vllm.model_executor.layers.fused_moe.fused_moe import (cutlass_moe_fp8,
+                                                            fused_experts,
+                                                            fused_topk)
+from vllm.utils import FlexibleArgumentParser
+
+DEFAULT_MODELS = [
+    "nm-testing/Mixtral-8x7B-Instruct-v0.1", "nm-testing/deepseekv2-lite",
+    "ibm-granite/granite-3.0-1b-a400m", "ibm-granite/granite-3.0-3b-a800m"
+]
+DEFAULT_BATCH_SIZES = [1, 4, 8, 16, 32, 64, 128, 256, 512]
+DEFAULT_TP_SIZES = [1]
+
+PER_ACT_TOKEN_OPTS = [False]
+PER_OUT_CH_OPTS = [False]
+
+
+def to_fp8(tensor: torch.Tensor):
+    finfo = torch.finfo(torch.float8_e4m3fn)
+    return torch.round(tensor.clamp(
+        min=finfo.min, max=finfo.max)).to(dtype=torch.float8_e4m3fn)
+
+
+def bench_run(results: list[benchmark.Measurement], model: str,
+              num_experts: int, topk: int, per_act_token: bool,
+              per_out_ch: bool, mkn: tuple[int, int, int]):
+    label = "Quant Matmul"
+
+    sub_label = (
+        "{}, num_experts={}, topk={}, per_act_token={} per_out_ch={}, "
+        "MKN=({})".format(model, num_experts, topk, per_act_token, per_out_ch,
+                          mkn))
+
+    print(f"Testing: {sub_label}")
+
+    (m, k, n) = mkn
+
+    dtype = torch.half
+
+    a = torch.randn((m, k), device="cuda", dtype=dtype) / 10
+    w1 = torch.randn((num_experts, 2 * n, k), device="cuda", dtype=dtype) / 10
+    w2 = torch.randn((num_experts, k, n), device="cuda", dtype=dtype) / 10
+
+    _, a_scale = ops.scaled_fp8_quant(a)
+
+    w1_q = torch.empty((num_experts, 2 * n, k),
+                       device="cuda",
+                       dtype=torch.float8_e4m3fn)
+    w2_q = torch.empty((num_experts, k, n),
+                       device="cuda",
+                       dtype=torch.float8_e4m3fn)
+    w1_scale = torch.empty((num_experts, 1, 1),
+                           device="cuda",
+                           dtype=torch.float32)
+    w2_scale = torch.empty((num_experts, 1, 1),
+                           device="cuda",
+                           dtype=torch.float32)
+
+    ab_strides1 = torch.full((num_experts, ),
+                             k,
+                             device="cuda",
+                             dtype=torch.int64)
+    c_strides1 = torch.full((num_experts, ),
+                            2 * n,
+                            device="cuda",
+                            dtype=torch.int64)
+    ab_strides2 = torch.full((num_experts, ),
+                             n,
+                             device="cuda",
+                             dtype=torch.int64)
+    c_strides2 = torch.full((num_experts, ),
+                            k,
+                            device="cuda",
+                            dtype=torch.int64)
+
+    for expert in range(num_experts):
+        w1_q[expert], w1_scale[expert] = ops.scaled_fp8_quant(w1[expert])
+        w2_q[expert], w2_scale[expert] = ops.scaled_fp8_quant(w2[expert])
+    w1_q_notransp = w1_q.clone()
+    w2_q_notransp = w2_q.clone()
+    w1_q = w1_q.transpose(1, 2)
+    w2_q = w2_q.transpose(1, 2)
+
+    score = torch.randn((m, num_experts), device="cuda", dtype=dtype)
+
+    topk_weights, topk_ids = fused_topk(a, score, topk, renormalize=False)
+
+    def run_triton_moe(a: torch.Tensor, w1: torch.Tensor, w2: torch.Tensor,
+                       topk_weights: torch.Tensor, topk_ids: torch.Tensor,
+                       w1_scale: torch.Tensor, w2_scale: torch.Tensor,
+                       a_scale: torch.Tensor, num_repeats: int):
+        for _ in range(num_repeats):
+            fused_experts(a,
+                          w1,
+                          w2,
+                          topk_weights,
+                          topk_ids,
+                          use_fp8_w8a8=True,
+                          w1_scale=w1_scale,
+                          w2_scale=w2_scale,
+                          a1_scale=a_scale)
+
+    def run_cutlass_moe(a: torch.Tensor, a_scale: torch.Tensor,
+                        w1: torch.Tensor, w2: torch.Tensor,
+                        w1_scale: torch.Tensor, w2_scale: torch.Tensor,
+                        topk_weights: torch.Tensor, topk_ids: torch.Tensor,
+                        ab_strides1: torch.Tensor, c_strides1: torch.Tensor,
+                        ab_strides2: torch.Tensor, c_strides2: torch.Tensor,
+                        num_repeats: int):
+        for _ in range(num_repeats):
+            cutlass_moe_fp8(a,
+                            w1,
+                            w2,
+                            w1_scale,
+                            w2_scale,
+                            topk_weights,
+                            topk_ids,
+                            ab_strides1,
+                            c_strides1,
+                            ab_strides2,
+                            c_strides2,
+                            a1_scale=a_scale)
+
+    def run_cutlass_from_graph(
+            a: torch.Tensor, a_scale: torch.Tensor, w1_q: torch.Tensor,
+            w2_q: torch.Tensor, w1_scale: torch.Tensor, w2_scale: torch.Tensor,
+            topk_weights: torch.Tensor, topk_ids: torch.Tensor,
+            ab_strides1: torch.Tensor, c_strides1: torch.Tensor,
+            ab_strides2: torch.Tensor, c_strides2: torch.Tensor):
+        with set_current_vllm_config(
+                VllmConfig(parallel_config=ParallelConfig(
+                    pipeline_parallel_size=1))):
+            return cutlass_moe_fp8(a,
+                                   w1_q,
+                                   w2_q,
+                                   w1_scale,
+                                   w2_scale,
+                                   topk_weights,
+                                   topk_ids,
+                                   ab_strides1,
+                                   c_strides1,
+                                   ab_strides2,
+                                   c_strides2,
+                                   a1_scale=a_scale)
+
+    def run_triton_from_graph(a: torch.Tensor, w1: torch.Tensor,
+                              w2: torch.Tensor, topk_weights: torch.Tensor,
+                              topk_ids: torch.Tensor, w1_scale: torch.Tensor,
+                              w2_scale: torch.Tensor, a_scale: torch.Tensor):
+        with set_current_vllm_config(
+                VllmConfig(parallel_config=ParallelConfig(
+                    pipeline_parallel_size=1))):
+            return fused_experts(a,
+                                 w1,
+                                 w2,
+                                 topk_weights,
+                                 topk_ids,
+                                 use_fp8_w8a8=True,
+                                 w1_scale=w1_scale,
+                                 w2_scale=w2_scale,
+                                 a1_scale=a_scale)
+
+    def replay_graph(graph, num_repeats):
+        for _ in range(num_repeats):
+            graph.replay()
+        torch.cuda.synchronize()
+
+    cutlass_stream = torch.cuda.Stream()
+    cutlass_graph = torch.cuda.CUDAGraph()
+    with torch.cuda.graph(cutlass_graph, stream=cutlass_stream):
+        run_cutlass_from_graph(a, a_scale, w1_q, w2_q, w1_scale, w2_scale,
+                               topk_weights, topk_ids, ab_strides1, c_strides1,
+                               ab_strides2, c_strides2)
+    torch.cuda.synchronize()
+
+    triton_stream = torch.cuda.Stream()
+    triton_graph = torch.cuda.CUDAGraph()
+    with torch.cuda.graph(triton_graph, stream=triton_stream):
+        run_triton_from_graph(a, w1_q_notransp, w2_q_notransp, topk_weights,
+                              topk_ids, w1_scale, w2_scale, a_scale)
+    torch.cuda.synchronize()
+
+    min_run_time = 5
+    num_warmup = 5
+    num_runs = 25
+
+    globals = {
+        # Baseline params
+        "w1": w1,
+        "w2": w2,
+        "score": score,
+        "topk": topk,
+        "w1_q_notransp": w1_q_notransp,
+        "w2_q_notransp": w2_q_notransp,
+        # Cutlass params
+        "a_scale": a_scale,
+        "w1_q": w1_q,
+        "w2_q": w2_q,
+        "w1_scale": w1_scale,
+        "w2_scale": w2_scale,
+        "ab_strides1": ab_strides1,
+        "c_strides1": c_strides1,
+        "ab_strides2": ab_strides2,
+        "c_strides2": c_strides2,
+        # cuda graph params
+        "cutlass_graph": cutlass_graph,
+        "triton_graph": triton_graph,
+        # Gen params
+        "a": a,
+        "topk_weights": topk_weights,
+        "topk_ids": topk_ids,
+        "num_runs": num_runs,
+        # Kernels
+        "run_triton_moe": run_triton_moe,
+        "run_cutlass_moe": run_cutlass_moe,
+        "replay_graph": replay_graph,
+    }
+
+    # Warmup
+    run_triton_moe(a, w1_q_notransp, w2_q_notransp, topk_weights, topk_ids,
+                   w1_scale, w2_scale, a_scale, num_warmup)
+
+    results.append(
+        benchmark.Timer(
+            stmt=
+            "run_triton_moe(a, w1_q_notransp, w2_q_notransp, topk_weights, topk_ids, w1_scale, w2_scale, a_scale, num_runs)",  # noqa: E501
+            globals=globals,
+            label=label,
+            sub_label=sub_label,
+            description="triton_moe",
+        ).blocked_autorange(min_run_time=min_run_time))
+
+    # Warmup
+    replay_graph(triton_graph, num_warmup)
+
+    results.append(
+        benchmark.Timer(
+            stmt="replay_graph(triton_graph, num_runs)",
+            globals=globals,
+            label=label,
+            sub_label=sub_label,
+            description="triton_moe_cuda_graphs",
+        ).blocked_autorange(min_run_time=min_run_time))
+
+    # Warmup
+    run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights,
+                    topk_ids, ab_strides1, c_strides1, ab_strides2, c_strides2,
+                    num_warmup)
+
+    results.append(
+        benchmark.Timer(
+            stmt=
+            "run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights, topk_ids, ab_strides1, c_strides1, ab_strides2, c_strides2, num_runs)",  # noqa: E501
+            globals=globals,
+            label=label,
+            sub_label=sub_label,
+            description="grouped_gemm_moe",
+        ).blocked_autorange(min_run_time=min_run_time))
+
+    # Warmup
+    replay_graph(cutlass_graph, num_warmup)
+
+    results.append(
+        benchmark.Timer(
+            stmt="replay_graph(cutlass_graph, num_runs)",
+            globals=globals,
+            label=label,
+            sub_label=sub_label,
+            description="grouped_gemm_moe_cuda_graphs",
+        ).blocked_autorange(min_run_time=min_run_time))
+
+
+def main(args):
+    print("Benchmarking models:")
+    for i, model in enumerate(args.models):
+        print(f"[{i}]  {model}")
+
+    results: list[benchmark.Measurement] = []
+
+    for model in args.models:
+        for tp in args.tp_sizes:
+            for layer in WEIGHT_SHAPES_MOE[model]:
+                num_experts = layer[0]
+                topk = layer[1]
+                size_k = layer[2]
+                size_n = layer[3] // tp
+
+                if len(args.limit_k) > 0 and size_k not in args.limit_k:
+                    continue
+
+                if len(args.limit_n) > 0 and size_n not in args.limit_n:
+                    continue
+
+                for per_act_token in PER_ACT_TOKEN_OPTS:
+                    for per_out_ch in PER_OUT_CH_OPTS:
+                        for size_m in DEFAULT_BATCH_SIZES:
+                            mkn = (size_m, size_k, size_n)
+                            bench_run(results, model, num_experts, topk,
+                                      per_act_token, per_out_ch, mkn)
+
+    compare = benchmark.Compare(results)
+    compare.print()
+
+
+if __name__ == "__main__":
+    parser = FlexibleArgumentParser(
+        description="Benchmark Marlin across specified models/shapes/batches")
+    parser.add_argument(
+        "--models",
+        nargs="+",
+        type=str,
+        default=DEFAULT_MODELS,
+        choices=WEIGHT_SHAPES_MOE.keys(),
+    )
+    parser.add_argument("--tp-sizes",
+                        nargs="+",
+                        type=int,
+                        default=DEFAULT_TP_SIZES)
+    parser.add_argument("--batch-sizes",
+                        nargs="+",
+                        type=int,
+                        default=DEFAULT_BATCH_SIZES)
+    parser.add_argument("--limit-k", nargs="+", type=int, default=[])
+    parser.add_argument("--limit-n", nargs="+", type=int, default=[])
+    parser.add_argument("--limit-num-groups", nargs="+", type=int, default=[])
+    parser.add_argument("--limit-per-act-token",
+                        nargs="+",
+                        type=int,
+                        default=[])
+    parser.add_argument("--limit-per-out-ch", nargs="+", type=int, default=[])
+
+    args = parser.parse_args()
+    main(args)

benchmarks/kernels/benchmark_paged_attention.py

@@ -7,10 +7,13 @@ from typing import Optional
 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 (STR_DTYPE_TO_TORCH_DTYPE, FlexibleArgumentParser,
                         create_kv_caches_with_random)
 
+logger = init_logger(__name__)
+
 NUM_BLOCKS = 128 * 1024
 PARTITION_SIZE = 512
 PARTITION_SIZE_ROCM = 256
@@ -193,6 +196,9 @@ def main(
 
 
 if __name__ == '__main__':
+    logger.warning("This script benchmarks the paged attention kernel. "
+                   "By default this is no longer used in vLLM inference.")
+
     parser = FlexibleArgumentParser(
         description="Benchmark the paged attention kernel.")
     parser.add_argument("--version",

benchmarks/kernels/benchmark_shapes.py

@@ -75,3 +75,19 @@ WEIGHT_SHAPES = {
         [7168, 8192],
     ],
 }
+
+WEIGHT_SHAPES_MOE = {
+    "nm-testing/Mixtral-8x7B-Instruct-v0.1": [
+        [8, 2, 4096, 28672],
+        [8, 2, 14336, 4096],
+    ],
+    "nm-testing/deepseekv2-lite": [
+        [64, 6, 2048, 1408],
+    ],
+    "ibm-granite/granite-3.0-1b-a400m": [
+        [32, 8, 1024, 1024],
+    ],
+    "ibm-granite/granite-3.0-3b-a800m": [
+        [40, 8, 1024, 1536],
+    ],
+}

benchmarks/kernels/benchmark_w8a8_block_fp8.py

@@ -0,0 +1,420 @@
+# SPDX-License-Identifier: Apache-2.0
+# Adapted from sglang quantization/tuning_block_wise_kernel.py
+
+import argparse
+import json
+import multiprocessing as mp
+import os
+import time
+from datetime import datetime
+from typing import Any
+
+import torch
+import tqdm
+import triton
+
+from vllm.model_executor.layers.quantization.utils.fp8_utils import (
+    _w8a8_block_fp8_matmul)
+from vllm.platforms import current_platform
+from vllm.utils import FlexibleArgumentParser
+
+mp.set_start_method("spawn", force=True)
+
+assert current_platform.is_cuda(
+), "Only support tune w8a8 block fp8 kernel on CUDA device."
+
+DTYPE_MAP = {
+    "float32": torch.float32,
+    "float16": torch.float16,
+    "half": torch.half,
+    "bfloat16": torch.bfloat16,
+}
+
+
+def w8a8_block_matmul(
+    A: torch.Tensor,
+    B: torch.Tensor,
+    As: torch.Tensor,
+    Bs: torch.Tensor,
+    block_size: list[int],
+    config: dict[str, Any],
+    output_dtype: torch.dtype = torch.float16,
+) -> torch.Tensor:
+    """This function performs matrix multiplication with 
+    block-wise quantization.
+
+    It takes two input tensors `A` and `B` with scales `As` and `Bs`.
+    The output is returned in the specified `output_dtype`.
+
+    Args:
+        A: The input tensor, e.g., activation.
+        B: The input tensor, e.g., weight.
+        As: The per-token-group quantization scale for `A`.
+        Bs: The per-block quantization scale for `B`.
+        block_size: The block size for per-block quantization. 
+                    It should be 2-dim, e.g., [128, 128].
+        output_dytpe: The dtype of the returned tensor.
+
+    Returns:
+        torch.Tensor: The result of matmul.
+    """
+    assert len(block_size) == 2
+    block_n, block_k = block_size[0], block_size[1]
+
+    assert A.shape[-1] == B.shape[-1]
+    assert A.shape[:-1] == As.shape[:-1] and A.is_contiguous()
+    assert triton.cdiv(A.shape[-1], block_k) == As.shape[-1]
+    M = A.numel() // A.shape[-1]
+
+    assert B.ndim == 2 and B.is_contiguous() and Bs.ndim == 2
+    N, K = B.shape
+    assert triton.cdiv(N, block_n) == Bs.shape[0]
+    assert triton.cdiv(K, block_k) == Bs.shape[1]
+
+    C_shape = A.shape[:-1] + (N, )
+    C = A.new_empty(C_shape, dtype=output_dtype)
+
+    def grid(META):
+        return (triton.cdiv(M, META["BLOCK_SIZE_M"]) *
+                triton.cdiv(N, META["BLOCK_SIZE_N"]), )
+
+    if A.dtype == torch.float8_e4m3fn:
+        kernel = _w8a8_block_fp8_matmul
+    else:
+        raise RuntimeError(
+            "Currently, only support tune w8a8 block fp8 kernel.")
+
+    kernel[grid](
+        A,
+        B,
+        C,
+        As,
+        Bs,
+        M,
+        N,
+        K,
+        block_n,
+        block_k,
+        A.stride(-2),
+        A.stride(-1),
+        B.stride(1),
+        B.stride(0),
+        C.stride(-2),
+        C.stride(-1),
+        As.stride(-2),
+        As.stride(-1),
+        Bs.stride(1),
+        Bs.stride(0),
+        **config,
+    )
+
+    return C
+
+
+def get_configs_compute_bound():
+    configs = []
+    for num_stages in [2, 3, 4, 5]:
+        for block_m in [16, 32, 64, 128, 256]:
+            for block_k in [64, 128]:
+                for block_n in [32, 64, 128, 256]:
+                    for num_warps in [4, 8]:
+                        for group_size in [1, 16, 32, 64]:
+                            configs.append({
+                                "BLOCK_SIZE_M": block_m,
+                                "BLOCK_SIZE_N": block_n,
+                                "BLOCK_SIZE_K": block_k,
+                                "GROUP_SIZE_M": group_size,
+                                "num_warps": num_warps,
+                                "num_stages": num_stages,
+                            })
+    return configs
+
+
+def get_weight_shapes(tp_size):
+    # NOTE(HandH1998): The weight shapes only works for DeepSeek-V3.
+    # Modify them, if you tune for another different model.
+    # cannot TP
+    total = [
+        (512 + 64, 7168),
+        ((128 + 64) * 128, 7168),
+        (128 * (128 + 128), 512),
+        (7168, 16384),
+        (7168, 18432),
+    ]
+    # N can TP
+    n_tp = [
+        (18432 * 2, 7168),
+        ((128 + 64) * 128, 7168),
+        (128 * (128 + 128), 512),
+        (24576, 1536),
+        (12288, 7168),
+        (4096, 7168),
+    ]
+    # K can TP
+    k_tp = [(7168, 18432), (7168, 16384), (7168, 2048)]
+
+    weight_shapes = []
+    for t in total:
+        weight_shapes.append(t)
+    for n_t in n_tp:
+        new_t = (n_t[0] // tp_size, n_t[1])
+        weight_shapes.append(new_t)
+    for k_t in k_tp:
+        new_t = (k_t[0], k_t[1] // tp_size)
+        weight_shapes.append(new_t)
+    return weight_shapes
+
+
+def benchmark_config(A,
+                     B,
+                     As,
+                     Bs,
+                     block_size,
+                     config,
+                     out_dtype=torch.float16,
+                     num_iters=10):
+
+    def run():
+        w8a8_block_matmul(A, B, As, Bs, block_size, config, out_dtype)
+
+    torch.cuda.synchronize()
+    # JIT complication & warmup
+    for _ in range(5):
+        run()
+    torch.cuda.synchronize()
+
+    start_event = torch.cuda.Event(enable_timing=True)
+    end_event = torch.cuda.Event(enable_timing=True)
+
+    latencies: list[float] = []
+    for i in range(num_iters):
+        torch.cuda.synchronize()
+        start_event.record()
+        run()
+        end_event.record()
+        end_event.synchronize()
+        latencies.append(start_event.elapsed_time(end_event))
+    avg = sum(latencies) / (num_iters * 10) * 1000  # us
+    return avg
+
+
+def tune(M, N, K, block_size, out_dtype, search_space, input_type):
+    factor_for_scale = 1e-2
+
+    if input_type == "fp8":
+        fp8_info = torch.finfo(torch.float8_e4m3fn)
+        fp8_max, fp8_min = fp8_info.max, fp8_info.min
+
+        A_fp32 = (
+            (torch.rand(M, K, dtype=torch.float32, device="cuda") - 0.5) * 2 *
+            fp8_max)
+        A = A_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn)
+
+        B_fp32 = (
+            (torch.rand(N, K, dtype=torch.float32, device="cuda") - 0.5) * 2 *
+            fp8_max)
+        B = B_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn)
+    else:
+        raise RuntimeError(
+            "Currently, only support tune w8a8 block fp8 kernel.")
+
+    block_n, block_k = block_size[0], block_size[1]
+    n_tiles = (N + block_n - 1) // block_n
+    k_tiles = (K + block_k - 1) // block_k
+
+    As = torch.rand(M, k_tiles, dtype=torch.float32,
+                    device="cuda") * factor_for_scale
+    Bs = (torch.rand(n_tiles, k_tiles, dtype=torch.float32, device="cuda") *
+          factor_for_scale)
+
+    best_config = None
+    best_time = float("inf")
+    for config in tqdm(search_space):
+        try:
+            kernel_time = benchmark_config(
+                A,
+                B,
+                As,
+                Bs,
+                block_size,
+                config,
+                out_dtype,
+                num_iters=10,
+            )
+        except triton.runtime.autotuner.OutOfResources:
+            # Some configurations may be invalid and fail to compile.
+            continue
+
+        if kernel_time < best_time:
+            best_time = kernel_time
+            best_config = config
+    now = datetime.now()
+    print(f"{now.ctime()}] Completed tuning for batch_size={M}")
+    assert best_config is not None
+    return best_config
+
+
+def save_configs(
+    N,
+    K,
+    block_n,
+    block_k,
+    configs,
+    save_path,
+    input_type="fp8",
+) -> None:
+    os.makedirs(save_path, exist_ok=True)
+    device_name = current_platform.get_device_name().replace(" ", "_")
+    json_file_name = (
+        f"N={N},K={K},device_name={device_name},dtype={input_type}_w8a8,"
+        f"block_shape=[{block_n},{block_k}].json")
+
+    config_file_path = os.path.join(save_path, json_file_name)
+    print(f"Writing best config to {config_file_path}...")
+
+    with open(config_file_path, "w") as f:
+        json.dump(configs, f, indent=4)
+        f.write("\n")
+
+
+def tune_on_gpu(args_dict):
+    """Run tuning on a specific GPU."""
+    gpu_id = args_dict["gpu_id"]
+    batch_sizes = args_dict["batch_sizes"]
+    weight_shapes = args_dict["weight_shapes"]
+    args = args_dict["args"]
+
+    torch.cuda.set_device(gpu_id)
+    print(f"Starting tuning on GPU {gpu_id} with batch sizes {batch_sizes}")
+
+    block_n = args.block_n
+    block_k = args.block_k
+    out_dtype = DTYPE_MAP[args.out_dtype]
+    save_path = args.save_path
+    input_type = args.input_type
+
+    search_space = get_configs_compute_bound()
+    search_space = [
+        config for config in search_space
+        if block_k % config["BLOCK_SIZE_K"] == 0
+    ]
+
+    start = time.time()
+    for shape in tqdm(weight_shapes, desc=f"GPU {gpu_id} - Shapes"):
+        N, K = shape[0], shape[1]
+        print(f"[GPU {gpu_id}] Tune for weight shape of `N: {N}, K: {K}`")
+        benchmark_results = [
+            tune(
+                batch_size,
+                N,
+                K,
+                [block_n, block_k],
+                out_dtype,
+                search_space,
+                input_type,
+            ) for batch_size in tqdm(batch_sizes,
+                                     desc=f"GPU {gpu_id} - Batch sizes")
+        ]
+        best_configs = {
+            M: config
+            for M, config in zip(batch_sizes, benchmark_results)
+        }
+        save_configs(N, K, block_n, block_k, best_configs, save_path,
+                     input_type)
+
+    end = time.time()
+    print(f"Tuning on GPU {gpu_id} took {end - start:.2f} seconds")
+
+
+def distribute_batch_sizes(batch_sizes, num_gpus):
+    """Distribute batch sizes across available GPUs."""
+    batches_per_gpu = []
+    for i in range(num_gpus):
+        start_idx = i * len(batch_sizes) // num_gpus
+        end_idx = (i + 1) * len(batch_sizes) // num_gpus
+        batches_per_gpu.append(batch_sizes[start_idx:end_idx])
+    return batches_per_gpu
+
+
+def main(args):
+    print(args)
+    num_gpus = torch.cuda.device_count()
+    if num_gpus == 0:
+        raise RuntimeError("No GPU available for tuning")
+    print(f"Found {num_gpus} GPUs for parallel tuning")
+
+    torch.cuda.init()
+
+    if args.batch_size is None:
+        batch_sizes = [
+            1,
+            2,
+            4,
+            8,
+            16,
+            24,
+            32,
+            48,
+            64,
+            96,
+            128,
+            256,
+            512,
+            1024,
+            1536,
+            2048,
+            3072,
+            4096,
+        ]
+    else:
+        batch_sizes = [args.batch_size]
+        num_gpus = 1  # If only one batch size, use only one GPU
+
+    weight_shapes = get_weight_shapes(args.tp_size)
+
+    batches_per_gpu = distribute_batch_sizes(batch_sizes, num_gpus)
+
+    process_args = []
+    for gpu_id in range(num_gpus):
+        process_args.append({
+            "gpu_id": gpu_id,
+            "batch_sizes": batches_per_gpu[gpu_id],
+            "weight_shapes":
+            weight_shapes,  # Each GPU processes all weight shapes
+            "args": args,
+        })
+
+    ctx = mp.get_context("spawn")
+    with ctx.Pool(num_gpus) as pool:
+        pool.map(tune_on_gpu, process_args)
+
+    print("Multi-GPU tuning completed")
+
+
+if __name__ == "__main__":
+    parser = FlexibleArgumentParser(
+        description="""
+Tune triton w8a8 block fp8 for DeepSeek-V3/DeepSeek-R1:
+    python3 benchmark_w8a8_block_fp8.py --tp-size 8 --input-type fp8
+Then copy to model_executor/layers/quantization/utils/configs
+        """,
+        formatter_class=argparse.RawTextHelpFormatter)
+
+    parser.add_argument("--tp-size", "-tp", type=int, default=8)
+    parser.add_argument("--input-type",
+                        type=str,
+                        choices=["fp8"],
+                        default="fp8")
+    parser.add_argument(
+        "--out-dtype",
+        type=str,
+        choices=["float32", "float16", "bfloat16", "half"],
+        default="float16",
+    )
+    parser.add_argument("--block-n", type=int, default=128)
+    parser.add_argument("--block-k", type=int, default=128)
+    parser.add_argument("--batch-size", type=int, required=False)
+    parser.add_argument("--save-path", type=str, default="./")
+    args = parser.parse_args()
+
+    main(args)

cmake/cpu_extension.cmake

@@ -190,6 +190,7 @@ set(VLLM_EXT_SRC
     "csrc/cpu/cache.cpp"
     "csrc/cpu/utils.cpp"
     "csrc/cpu/layernorm.cpp"
+    "csrc/cpu/mla_decode.cpp"
     "csrc/cpu/pos_encoding.cpp"
     "csrc/cpu/torch_bindings.cpp")
 

csrc/cpu/cache.cpp

@@ -88,6 +88,48 @@ void reshape_and_cache_cpu_impl(
 }
 };  // namespace
 
+template <typename scalar_t>
+void concat_and_cache_mla_cpu_impl(
+    const scalar_t* __restrict__ kv_c,  // [num_tokens, kv_lora_rank]
+    const scalar_t* __restrict__ k_pe,  // [num_tokens, pe_dim]
+    scalar_t* __restrict__ kv_cache,  // [num_blocks, block_size, (kv_lora_rank
+                                      // + pe_dim)]
+    const int64_t* __restrict__ slot_mapping,  // [num_tokens]
+    const int num_tokens,                      //
+    const int block_stride,                    //
+    const int entry_stride,                    //
+    const int kv_c_stride,                     //
+    const int k_pe_stride,                     //
+    const int kv_lora_rank,                    //
+    const int pe_dim,                          //
+    const int block_size                       //
+) {
+#pragma omp parallel for
+  for (int token_idx = 0; token_idx < num_tokens; ++token_idx) {
+    const int64_t slot_idx = slot_mapping[token_idx];
+    // NOTE: slot_idx can be -1 if the token is padded
+    if (slot_idx < 0) {
+      continue;
+    }
+    const int64_t block_idx = slot_idx / block_size;
+    const int64_t block_offset = slot_idx % block_size;
+
+    auto copy = [&](const scalar_t* __restrict__ src,
+                    scalar_t* __restrict__ dst, int src_stride, int dst_stride,
+                    int size, int offset) {
+      for (int i = 0; i < size; i++) {
+        const int64_t src_idx = token_idx * src_stride + i;
+        const int64_t dst_idx =
+            block_idx * block_stride + block_offset * entry_stride + i + offset;
+        dst[dst_idx] = src[src_idx];
+      }
+    };
+
+    copy(kv_c, kv_cache, kv_c_stride, block_stride, kv_lora_rank, 0);
+    copy(k_pe, kv_cache, k_pe_stride, block_stride, pe_dim, kv_lora_rank);
+  }
+}
+
 // Note: the key_caches and value_caches vectors are constant but
 // not the Tensors they contain. The vectors need to be const refs
 // in order to satisfy pytorch's C++ operator registration code.
@@ -134,6 +176,38 @@ void reshape_and_cache(torch::Tensor& key, torch::Tensor& value,
   });
 }
 
+void concat_and_cache_mla(
+    torch::Tensor& kv_c,          // [num_tokens, kv_lora_rank]
+    torch::Tensor& k_pe,          // [num_tokens, pe_dim]
+    torch::Tensor& kv_cache,      // [num_blocks, block_size, (kv_lora_rank +
+                                  // pe_dim)]
+    torch::Tensor& slot_mapping,  // [num_tokens] or [num_actual_tokens]
+    const std::string& kv_cache_dtype, torch::Tensor& scale) {
+  int num_tokens = slot_mapping.size(0);
+  int kv_lora_rank = kv_c.size(1);
+  int pe_dim = k_pe.size(1);
+  int block_size = kv_cache.size(1);
+
+  TORCH_CHECK(kv_cache.size(2) == kv_lora_rank + pe_dim);
+  TORCH_CHECK(kv_cache_dtype != "fp8");
+
+  int kv_c_stride = kv_c.stride(0);
+  int k_pe_stride = k_pe.stride(0);
+  int block_stride = kv_cache.stride(0);
+  int entry_stride = kv_cache.stride(1);
+
+  VLLM_DISPATCH_FLOATING_TYPES(
+      kv_c.scalar_type(), "concat_and_cache_mla_cpu_impl", [&] {
+        CPU_KERNEL_GUARD_IN(concat_and_cache_mla_cpu_impl)
+        concat_and_cache_mla_cpu_impl<scalar_t>(
+            kv_c.data_ptr<scalar_t>(), k_pe.data_ptr<scalar_t>(),
+            kv_cache.data_ptr<scalar_t>(), slot_mapping.data_ptr<int64_t>(),
+            num_tokens, block_stride, entry_stride, kv_c_stride, k_pe_stride,
+            kv_lora_rank, pe_dim, block_size);
+        CPU_KERNEL_GUARD_OUT(concat_and_cache_mla_cpu_impl)
+      });
+}
+
 void swap_blocks(torch::Tensor& src, torch::Tensor& dst,
                  const torch::Tensor& block_mapping) {
   TORCH_CHECK(false, "swap_blocks is unsupported on CPU.")

csrc/cpu/cpu_types_x86.hpp

@@ -130,6 +130,8 @@ struct BF16Vec32 : public Vec<BF16Vec32> {
 
   __m512i reg;
 
+  explicit BF16Vec32() : reg(_mm512_setzero_si512()) {}
+
   explicit BF16Vec32(const void* ptr) : reg((__m512i)_mm512_loadu_si512(ptr)) {}
 
   explicit BF16Vec32(__m512i data) : reg(data) {}

csrc/cpu/mla_decode.cpp

@@ -0,0 +1,393 @@
+#include "cpu_types.hpp"
+#include <float.h>
+
+namespace {
+template <typename scalar_t>
+struct KernelVecType {
+  using qk_load_vec_type = void;
+  using qk_vec_type = void;
+  using v_load_vec_type = void;
+};
+
+template <>
+struct KernelVecType<float> {
+  using qk_load_vec_type = vec_op::FP32Vec16;
+  using qk_vec_type = vec_op::FP32Vec16;
+  using v_load_vec_type = vec_op::FP32Vec16;
+};
+
+template <>
+struct KernelVecType<c10::Half> {
+#if defined(__powerpc64__) || defined(__s390x__)
+  // Power and s390x architecture-specific vector types
+  using qk_load_vec_type = vec_op::FP32Vec16;
+  using qk_vec_type = vec_op::FP32Vec16;
+  using v_load_vec_type = vec_op::FP32Vec16;
+#else
+  // Fallback for other architectures, including x86
+  using qk_load_vec_type = vec_op::FP16Vec16;
+  using qk_vec_type = vec_op::FP32Vec16;
+  using v_load_vec_type = vec_op::FP16Vec16;
+#endif
+};
+
+#ifdef __AVX512BF16__
+template <>
+struct KernelVecType<c10::BFloat16> {
+  using qk_load_vec_type = vec_op::BF16Vec32;
+  using qk_vec_type = vec_op::BF16Vec32;
+  using v_load_vec_type = vec_op::BF16Vec16;
+};
+#elif defined(__aarch64__) && !defined(ARM_BF16_SUPPORT)
+// pass
+#else
+template <>
+struct KernelVecType<c10::BFloat16> {
+  using qk_load_vec_type = vec_op::BF16Vec16;
+  using qk_vec_type = vec_op::FP32Vec16;
+  using v_load_vec_type = vec_op::BF16Vec16;
+};
+#endif
+
+template <int HEAD_DIM, int V_HEAD_DIM, int BLOCK_SIZE, int HEAD_UNROLL,
+          typename qk_vec_type>
+void mla_decode_block_head(
+    const qk_vec_type* __restrict__ q_vecs,          // [HEAD_UNROLL, head_dim]
+    const qk_vec_type* __restrict__ k_vecs,          // [block_size, head_dim]
+    const vec_op::FP32Vec16* __restrict v_vecs_f32,  // [block_size, v_head_dim]
+    float* __restrict__ acc_out,  // [HEAD_UNROLL, v_head_dim]
+    float* __restrict__ acc_lse,  // [HEAD_UNROLL]
+    const float scale, const int num_tokens) {
+  using f32_vec_type = vec_op::FP32Vec16;
+  constexpr int QK_NUM_ELEM = qk_vec_type::VEC_ELEM_NUM;
+  constexpr int V_NUM_ELEM = f32_vec_type::VEC_ELEM_NUM;
+
+  float logits[BLOCK_SIZE][HEAD_UNROLL] = {};  // initialize to zeros
+  float max_val[HEAD_UNROLL];
+  std::fill(max_val, max_val + HEAD_UNROLL, -FLT_MAX);
+
+  f32_vec_type acc_vec[BLOCK_SIZE][HEAD_UNROLL];
+  for (int i = 0; i < HEAD_DIM; i += QK_NUM_ELEM) {
+    // load to registers
+    qk_vec_type q_vec[HEAD_UNROLL];
+
+#pragma unroll
+    for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll)
+      q_vec[unroll] =
+          qk_vec_type{q_vecs[(i + unroll * HEAD_DIM) / QK_NUM_ELEM]};
+
+    for (int block_offset = 0; block_offset < num_tokens; ++block_offset) {
+      qk_vec_type k_vec(k_vecs[(block_offset * HEAD_DIM + i) / QK_NUM_ELEM]);
+
+#pragma unroll
+      for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll)
+        vec_op::fma(acc_vec[block_offset][unroll], q_vec[unroll], k_vec);
+    }
+  }
+
+  for (int block_offset = 0; block_offset < num_tokens; ++block_offset) {
+#pragma unroll
+    for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll) {
+      const float acc = acc_vec[block_offset][unroll].reduce_sum() * scale;
+      logits[block_offset][unroll] = acc;
+      max_val[unroll] = std::max(max_val[unroll], acc);
+    }
+  }
+
+  float sum_exp[HEAD_UNROLL] = {};
+  for (int block_offset = 0; block_offset < num_tokens; ++block_offset) {
+#pragma unroll
+    for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll) {
+      const float val =
+          std::exp(logits[block_offset][unroll] - max_val[unroll]);
+      logits[block_offset][unroll] = val;
+      sum_exp[unroll] += val;
+    }
+  }
+
+  f32_vec_type this_out[V_HEAD_DIM / V_NUM_ELEM][HEAD_UNROLL];
+
+  for (int block_offset = 0; block_offset < num_tokens; ++block_offset) {
+    // load to registers
+    f32_vec_type scale_[HEAD_UNROLL];
+
+#pragma unroll
+    for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll)
+      scale_[unroll] =
+          f32_vec_type{logits[block_offset][unroll] / sum_exp[unroll]};
+
+    for (int i = 0; i < V_HEAD_DIM; i += V_NUM_ELEM) {
+      f32_vec_type v_vec(
+          v_vecs_f32[(block_offset * HEAD_DIM + i) / V_NUM_ELEM]);
+
+#pragma unroll
+      for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll)
+        vec_op::fma(this_out[i / V_NUM_ELEM][unroll], v_vec, scale_[unroll]);
+    }
+  }
+
+  // merge attention state
+  // section 2.2 in https://arxiv.org/pdf/2501.01005
+  f32_vec_type prev_scale[HEAD_UNROLL];
+  f32_vec_type curr_scale[HEAD_UNROLL];
+
+#pragma unroll
+  for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll) {
+    const float prev_lse = acc_lse[unroll];
+    const float curr_lse = std::log(sum_exp[unroll]) +
+                           max_val[unroll];  // add back max_val to get true lse
+    // softmax trick
+    const float max_lse = std::max(prev_lse, curr_lse);
+    const float prev_sum_exp = std::exp(prev_lse - max_lse);
+    const float curr_sum_exp = std::exp(curr_lse - max_lse);
+
+    const float new_sum_exp = prev_sum_exp + curr_sum_exp;
+    acc_lse[unroll] = std::log(new_sum_exp) + max_lse;
+
+    prev_scale[unroll] = f32_vec_type{prev_sum_exp / new_sum_exp};
+    curr_scale[unroll] = f32_vec_type{curr_sum_exp / new_sum_exp};
+  }
+
+  for (int i = 0; i < V_HEAD_DIM; i += V_NUM_ELEM) {
+#pragma unroll
+    for (int unroll = 0; unroll < HEAD_UNROLL; ++unroll) {
+      f32_vec_type o_vec(acc_out + i + V_HEAD_DIM * unroll);
+      o_vec = o_vec * prev_scale[unroll] +
+              this_out[i / V_NUM_ELEM][unroll] * curr_scale[unroll];
+      o_vec.save(acc_out + i + V_HEAD_DIM * unroll);
+    }
+  }
+
+  q_vecs += HEAD_DIM / QK_NUM_ELEM * HEAD_UNROLL;
+  acc_out += V_HEAD_DIM * HEAD_UNROLL;
+}
+
+template <typename scalar_t, int HEAD_DIM, int V_HEAD_DIM, int BLOCK_SIZE,
+          typename qk_vec_type>
+void mla_decode_block(
+    const qk_vec_type* __restrict__ q_vecs,  // [num_heads, head_dim]
+    const scalar_t* __restrict__ kv_cache,   // [block_size, head_dim]
+    float* __restrict__ acc_out,             // [num_heads, v_head_dim]
+    float* __restrict__ acc_lse,             // [num_heads]
+    const int num_heads, const float scale, const int num_tokens) {
+  using qk_load_vec_type = typename KernelVecType<scalar_t>::qk_load_vec_type;
+  static_assert(
+      std::is_same<qk_vec_type,
+                   typename KernelVecType<scalar_t>::qk_vec_type>::value);
+  using v_load_vec_type = typename KernelVecType<scalar_t>::v_load_vec_type;
+  using f32_vec_type = vec_op::FP32Vec16;
+  static_assert(qk_load_vec_type::VEC_ELEM_NUM == qk_vec_type::VEC_ELEM_NUM);
+  static_assert(v_load_vec_type::VEC_ELEM_NUM == f32_vec_type::VEC_ELEM_NUM);
+  constexpr int QK_NUM_ELEM = qk_vec_type::VEC_ELEM_NUM;
+  constexpr int V_NUM_ELEM = v_load_vec_type::VEC_ELEM_NUM;
+
+  const qk_vec_type* k_vecs;
+  const f32_vec_type* v_vecs_f32;
+  float* kv_cache_f32 = nullptr;
+
+  if constexpr (!std::is_same<scalar_t, float>::value) {
+    // convert KV cache block to FP32 to reuse it across query heads and
+    // attn @ V computation, since FP16/BF16->FP32 is expensive.
+    // TODO: move malloc outside of this fn to reuse across iterations.
+    const int nbytes = BLOCK_SIZE * HEAD_DIM * sizeof(float);
+    kv_cache_f32 = static_cast<float*>(std::aligned_alloc(64, nbytes));
+
+    for (int block_offset = 0; block_offset < num_tokens; ++block_offset)
+      for (int i = 0; i < HEAD_DIM; i += V_NUM_ELEM) {
+        v_load_vec_type kv_load_vec(kv_cache + block_offset * HEAD_DIM + i);
+        f32_vec_type kv_vec_f32(kv_load_vec);
+        kv_vec_f32.save(kv_cache_f32 + block_offset * HEAD_DIM + i);
+      }
+
+    if constexpr (std::is_same<qk_load_vec_type, qk_vec_type>::value) {
+      // for AVX512_BF16, Q @ K.T uses BF16 for K (no conversion)
+      // NOTE: in this case, we only need to convert the V section to FP32.
+      // But for simplicity, we will convert the whole KV block to FP32.
+      k_vecs = reinterpret_cast<const qk_vec_type*>(kv_cache);
+    } else {
+      k_vecs = reinterpret_cast<const qk_vec_type*>(kv_cache_f32);
+    }
+
+    // attn @ V always use FP32 for V, since attn is FP32.
+    v_vecs_f32 = reinterpret_cast<const f32_vec_type*>(kv_cache_f32);
+
+  } else {
+    // KV cache is FP32. don't need to do anything.
+    k_vecs = reinterpret_cast<const qk_vec_type*>(kv_cache);
+    v_vecs_f32 = reinterpret_cast<const f32_vec_type*>(kv_cache);
+  }
+
+  // compute 2 heads at the same time to improve ILP and
+  // take advantage of register cache for K and V.
+  constexpr int HEAD_UNROLL = 2;
+  for (int iter = 0; iter < num_heads / HEAD_UNROLL; ++iter) {
+    mla_decode_block_head<HEAD_DIM, V_HEAD_DIM, BLOCK_SIZE, HEAD_UNROLL>(
+        q_vecs, k_vecs, v_vecs_f32, acc_out, acc_lse, scale, num_tokens);
+
+    q_vecs += HEAD_UNROLL * HEAD_DIM / QK_NUM_ELEM;
+    acc_out += HEAD_UNROLL * V_HEAD_DIM;
+    acc_lse += HEAD_UNROLL;
+  }
+
+  // take care of the remaining heads
+  for (int iter = 0; iter < num_heads % HEAD_UNROLL; ++iter) {
+    mla_decode_block_head<HEAD_DIM, V_HEAD_DIM, BLOCK_SIZE, 1>(
+        q_vecs, k_vecs, v_vecs_f32, acc_out, acc_lse, scale, num_tokens);
+
+    q_vecs += HEAD_DIM / QK_NUM_ELEM;
+    acc_out += V_HEAD_DIM;
+    acc_lse += 1;
+  }
+
+  if (kv_cache_f32 != nullptr) {
+    std::free(kv_cache_f32);
+  }
+}
+}  // namespace
+
+template <typename scalar_t, int HEAD_DIM, int V_HEAD_DIM, int BLOCK_SIZE>
+void mla_decode_kvcache_cpu_impl(
+    scalar_t* __restrict__ out,             // [num_seqs, num_heads, v_head_dim]
+    const scalar_t* __restrict__ q,         // [num_seqs, num_heads, head_dim]
+    const scalar_t* __restrict__ kv_cache,  // [num_blocks, block_size,
+                                            // head_dim]
+    const int num_heads, const float scale,
+    const int* __restrict__ block_tables,  // [num_seqs, max_num_blocks_per_seq]
+    const int* __restrict__ seq_lens,      // [num_seqs]
+    const int max_num_blocks_per_seq, const int o_stride, const int q_stride,
+    const int kv_stride, const int num_seqs) {
+  using qk_load_vec_type = typename KernelVecType<scalar_t>::qk_load_vec_type;
+  using qk_vec_type = typename KernelVecType<scalar_t>::qk_vec_type;
+  constexpr int QK_NUM_ELEM = qk_vec_type::VEC_ELEM_NUM;
+
+  // shared across threads
+  const int max_threads = omp_get_max_threads();
+  const int acc_out_nbytes =
+      max_threads * num_heads * V_HEAD_DIM * sizeof(float);
+  float* acc_out = static_cast<float*>(std::aligned_alloc(64, acc_out_nbytes));
+  std::vector<float> acc_lse(max_threads * num_heads);
+
+  // allocate memory to pre-convert query to FP32 later
+  float* q_f32;
+  constexpr bool PRE_CONVERT_QUERY =
+      !std::is_same<scalar_t, float>::value &&
+      std::is_same<qk_vec_type, vec_op::FP32Vec16>::value;
+  if constexpr (PRE_CONVERT_QUERY) {
+    const int q_f32_nbytes = num_heads * HEAD_DIM * sizeof(float);
+    q_f32 = static_cast<float*>(std::aligned_alloc(64, q_f32_nbytes));
+  }
+
+#pragma omp parallel
+  {
+    const int num_threads = omp_get_num_threads();
+    const int thread_id = omp_get_thread_num();
+    float* __restrict__ acc_out_thread =
+        acc_out + thread_id * num_heads * V_HEAD_DIM;
+    float* __restrict__ acc_lse_thread = acc_lse.data() + thread_id * num_heads;
+
+    for (int seq_idx = 0; seq_idx < num_seqs; ++seq_idx) {
+      // reset accumulator
+      std::fill(acc_out_thread, acc_out_thread + num_heads * V_HEAD_DIM, 0.0f);
+      std::fill(acc_lse_thread, acc_lse_thread + num_heads, -FLT_MAX);
+
+      const int seq_len = seq_lens[seq_idx];
+      const int block_num = (seq_len + BLOCK_SIZE - 1) / BLOCK_SIZE;
+      const int last_block_size = seq_len - (block_num - 1) * BLOCK_SIZE;
+
+      const qk_vec_type* q_vecs;
+      if constexpr (PRE_CONVERT_QUERY) {
+// pre-convert query to FP32 since FP16/BF16->FP32 is slow.
+#pragma omp for
+        for (int i = 0; i < num_heads * HEAD_DIM; i += QK_NUM_ELEM) {
+          qk_load_vec_type q_load_vec(q + seq_idx * q_stride + i);
+          qk_vec_type q_vec(q_load_vec);
+          q_vec.save(q_f32 + i);
+        }
+        q_vecs = reinterpret_cast<const qk_vec_type*>(q_f32);
+      } else {
+        q_vecs = reinterpret_cast<const qk_vec_type*>(q + seq_idx * q_stride);
+      }
+
+#pragma omp for
+      for (int block_idx = 0; block_idx < block_num; ++block_idx) {
+        const int physical_block_idx =
+            block_tables[seq_idx * max_num_blocks_per_seq + block_idx];
+        const int num_tokens =
+            block_idx < block_num - 1 ? BLOCK_SIZE : last_block_size;
+
+        mla_decode_block<scalar_t, HEAD_DIM, V_HEAD_DIM, BLOCK_SIZE>(
+            q_vecs, kv_cache + physical_block_idx * kv_stride, acc_out_thread,
+            acc_lse_thread, num_heads, scale, num_tokens);
+      }
+
+// merge attention states across threads
+// section 2.2 in https://arxiv.org/pdf/2501.01005
+// each thread is responsible for 1 head
+#pragma omp for
+      for (int head_idx = 0; head_idx < num_heads; ++head_idx) {
+        float* acc_lse_head = acc_lse.data() + head_idx;
+        float* acc_out_head = acc_out + head_idx * V_HEAD_DIM;
+
+        float max_val = -FLT_MAX;
+        for (int thread_id_ = 0; thread_id_ < num_threads; ++thread_id_) {
+          max_val = std::max(max_val, acc_lse_head[thread_id_ * num_heads]);
+        }
+
+        float sum_exp = 0.0f;
+        for (int thread_id_ = 0; thread_id_ < num_threads; ++thread_id_) {
+          float val = std::exp(acc_lse_head[thread_id_ * num_heads] - max_val);
+          acc_lse_head[thread_id_ * num_heads] = val;
+          sum_exp += val;
+        }
+
+        float inv_sum = 1.0f / sum_exp;
+        float out_head[V_HEAD_DIM] = {};
+        for (int thread_id_ = 0; thread_id_ < num_threads; ++thread_id_) {
+          float scale_ = acc_lse_head[thread_id_ * num_heads] * inv_sum;
+          for (int i = 0; i < V_HEAD_DIM; ++i) {
+            out_head[i] +=
+                acc_out_head[thread_id_ * num_heads * V_HEAD_DIM + i] * scale_;
+          }
+        }
+
+        for (int i = 0; i < V_HEAD_DIM; ++i) {
+          vec_op::storeFP32(out_head[i], out + seq_idx * o_stride +
+                                             head_idx * V_HEAD_DIM + i);
+        }
+      }
+    }
+  }
+  if (PRE_CONVERT_QUERY) {
+    std::free(q_f32);
+  }
+  std::free(acc_out);
+}
+
+void mla_decode_kvcache(torch::Tensor& out, torch::Tensor& query,
+                        torch::Tensor& kv_cache, double scale,
+                        torch::Tensor& block_tables, torch::Tensor& seq_lens) {
+  const int num_seqs = query.size(0);
+  const int num_heads = query.size(1);
+  const int head_dim = query.size(2);
+  const int block_size = kv_cache.size(1);
+  const int v_head_dim = out.size(2);
+
+  const int max_num_blocks_per_seq = block_tables.size(1);
+  const int o_stride = out.stride(0);
+  const int q_stride = query.stride(0);
+  const int kv_stride = kv_cache.stride(0);
+
+  VLLM_DISPATCH_FLOATING_TYPES(
+      query.scalar_type(), "mla_decode_kvcache_cpu_impl", [&] {
+        CPU_KERNEL_GUARD_IN(mla_decode_kvcache_cpu_impl)
+        if (head_dim == 576 && v_head_dim == 512 && block_size == 16)
+          mla_decode_kvcache_cpu_impl<scalar_t, 576, 512, 16>(
+              out.data_ptr<scalar_t>(), query.data_ptr<scalar_t>(),
+              kv_cache.data_ptr<scalar_t>(), num_heads, scale,
+              block_tables.data_ptr<int>(), seq_lens.data_ptr<int>(),
+              max_num_blocks_per_seq, o_stride, q_stride, kv_stride, num_seqs);
+        else
+          TORCH_CHECK(false, "Unsupported block size: ", block_size);
+        CPU_KERNEL_GUARD_OUT(mla_decode_kvcache_cpu_impl)
+      });
+}

csrc/cpu/torch_bindings.cpp

@@ -18,6 +18,10 @@ void int8_scaled_mm_azp(torch::Tensor& c, const torch::Tensor& a,
                         const std::optional<torch::Tensor>& azp,
                         const std::optional<torch::Tensor>& bias);
 
+void mla_decode_kvcache(torch::Tensor& out, torch::Tensor& query,
+                        torch::Tensor& kv_cache, double scale,
+                        torch::Tensor& block_tables, torch::Tensor& seq_lens);
+
 TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   // vLLM custom ops
 
@@ -150,6 +154,14 @@ TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _cache_ops), cache_ops) {
       "                  str kv_cache_dtype,"
       "                  Tensor k_scale, Tensor v_scale) -> ()");
   cache_ops.impl("reshape_and_cache", torch::kCPU, &reshape_and_cache);
+
+  cache_ops.def(
+      "concat_and_cache_mla(Tensor kv_c, Tensor k_pe,"
+      "                     Tensor! kv_cache,"
+      "                     Tensor slot_mapping,"
+      "                     str kv_cache_dtype,"
+      "                     Tensor scale) -> ()");
+  cache_ops.impl("concat_and_cache_mla", torch::kCPU, &concat_and_cache_mla);
 }
 
 TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _utils), utils) {
@@ -157,4 +169,12 @@ TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _utils), utils) {
   utils.def("init_cpu_threads_env(str cpu_ids) -> str", &init_cpu_threads_env);
 }
 
+TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _cpu), cpu_ops) {
+  cpu_ops.def(
+      "mla_decode_kvcache("
+      "   Tensor! out, Tensor query, Tensor kv_cache,"
+      "   float scale, Tensor block_tables, Tensor seq_lens) -> ()");
+  cpu_ops.impl("mla_decode_kvcache", torch::kCPU, &mla_decode_kvcache);
+}
+
 REGISTER_EXTENSION(TORCH_EXTENSION_NAME)

csrc/cutlass_extensions/common.hpp

@@ -48,4 +48,14 @@ struct enable_sm90_or_later : Kernel {
     Kernel::operator()(std::forward<Args>(args)...);
 #endif
   }
-};
+};
+
+template <typename Kernel>
+struct enable_sm90_only : Kernel {
+  template <typename... Args>
+  CUTLASS_DEVICE void operator()(Args&&... args) {
+#if defined __CUDA_ARCH__ && __CUDA_ARCH__ == 900
+    Kernel::operator()(std::forward<Args>(args)...);
+#endif
+  }
+};

csrc/cutlass_extensions/epilogue/broadcast_load_epilogue_array_c3x.hpp

@@ -0,0 +1,457 @@
+/***************************************************************************************************
+ * Copyright (c) 2023 - 2024 NVIDIA CORPORATION & AFFILIATES. All rights
+ *reserved. SPDX-License-Identifier: BSD-3-Clause
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ *ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ *LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ *CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ *SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ *INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ *CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ *ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ *POSSIBILITY OF SUCH DAMAGE.
+ *
+ **************************************************************************************************/
+
+//
+// This file is a modified excerpt of
+// include/cutlass/epilogue/fusion/sm90_visitor_load_tma_warpspecialized.hpp
+// from https://github.com/NVIDIA/cutlass v3.5.0
+// It has been modified to support either row/column or scalar broadcasting
+// where the tensor being loaded from is always passed in via a device pointer.
+// This lets one compiled kernel handle all cases of per-tensor or
+// per-channel/per-token quantization.
+//
+// This interface also allows the scales to be passed in as tensors that
+// consistently reside on the device, which avoids an issue with a previous
+// implementation where scalars needed to be on the CPU since they
+// were passed in via float values. This created a potential performance hazard
+// if scales were initially on the device, and caused torch.compile graphs
+// breaks when moving scales to the CPU.
+//
+#pragma once
+
+// Turn off clang-format for the entire file to keep it close to upstream
+// clang-format off
+
+#include "cutlass/cutlass.h"
+#include "cutlass/arch/barrier.h"
+
+#include "cute/tensor.hpp"
+#include "cutlass/epilogue/fusion/sm90_visitor_tma_warpspecialized.hpp"
+
+namespace cutlass::epilogue::fusion {
+
+using namespace cute;
+using namespace detail;
+
+// Row vector broadcast
+template<
+  int Stages,
+  class CtaTileShapeMNK,
+  class Element,
+  class StrideMNL = Stride<_0,_1,_0>,
+  int Alignment = 128 / sizeof_bits_v<Element>
+>
+struct Sm90RowOrScalarBroadcastArray {
+  static_assert(Stages == 0, "Row broadcast doesn't support smem usage");
+  static_assert(is_static_v<decltype(take<0,2>(StrideMNL{}))>); // batch stride can be dynamic or static
+  static_assert(take<0,2>(StrideMNL{}) == Stride<_0,_1>{});
+
+  struct SharedStorage { 
+    array_aligned<Element, size<1>(CtaTileShapeMNK{})> smem;
+  };
+
+  // This struct has been modified to have a bool indicating that ptr_row is a 
+  // scalar that must be broadcast, instead of containing a scalar that is 
+  // valid if ptr_row is null.
+  struct Arguments {
+    const Element* const* ptr_row_array = nullptr;
+    bool row_broadcast = true;
+    StrideMNL dRow = {};
+  };
+
+  using Params = Arguments;
+
+  template <class ProblemShape>
+  static constexpr Params
+  to_underlying_arguments(ProblemShape const& problem_shape, Arguments const& args, void* workspace) {
+    return args;
+  }
+
+  template <class ProblemShape>
+  static bool
+  can_implement(ProblemShape const& problem_shape, Arguments const& args) {
+    return true;
+  }
+
+  template <class ProblemShape>
+  static size_t
+  get_workspace_size(ProblemShape const& problem_shape, Arguments const& args) {
+    return 0;
+  }
+
+  template <class ProblemShape>
+  static cutlass::Status
+  initialize_workspace(ProblemShape const& problem_shape, Arguments const& args, void* workspace, cudaStream_t stream,
+    CudaHostAdapter* cuda_adapter = nullptr) {
+    return cutlass::Status::kSuccess;
+  }
+
+  CUTLASS_HOST_DEVICE
+  Sm90RowOrScalarBroadcastArray() { }
+
+  CUTLASS_HOST_DEVICE
+  Sm90RowOrScalarBroadcastArray(Params const& params, SharedStorage const& shared_storage)
+      : params(params)
+      , smem(const_cast<Element*>(shared_storage.smem.data())) { }
+
+  Params params;
+  Element *smem = nullptr;
+
+  CUTLASS_DEVICE bool
+  is_producer_load_needed() const {
+    return false;
+  }
+
+  CUTLASS_DEVICE bool
+  is_C_load_needed() const {
+    return false;
+  }
+
+  CUTLASS_DEVICE bool
+  is_zero() const {
+    return (!params.row_broadcast && *(params.ptr_row_array[group]) == Element(0));
+  }
+
+  template <class... Args>
+  CUTLASS_DEVICE auto
+  get_producer_load_callbacks(ProducerLoadArgs<Args...> const& args) {
+    return EmptyProducerLoadCallbacks{};
+  }
+
+  template <class GS_GTensor, class GS_STensor, class GS_CTensor, class Tiled_G2S, class SR_STensor, class SR_RTensor, class CTensor, class ThrResidue, class ThrNum>
+  struct ConsumerStoreCallbacks : EmptyConsumerStoreCallbacks {
+    CUTLASS_DEVICE
+    ConsumerStoreCallbacks(
+        GS_GTensor tGS_gRow_, GS_STensor tGS_sRow_, 
+        GS_CTensor tGS_cRow_, Tiled_G2S tiled_g2s_, 
+        SR_STensor tSR_sRow_, SR_RTensor tSR_rRow_,
+        CTensor tCcRow_, ThrResidue residue_tCcRow_, ThrNum thr_num_,
+        int group, Params const& params_)
+      : tGS_gRow(tGS_gRow_)
+      , tGS_sRow(tGS_sRow_)
+      , tGS_cRow(tGS_cRow_)
+      , tiled_G2S(tiled_g2s_)
+      , tSR_sRow(tSR_sRow_)
+      , tSR_rRow(tSR_rRow_)
+      , tCcRow(tCcRow_)
+      , residue_tCcRow(residue_tCcRow_)
+      , group(group)
+      , params(params_) {}
+
+    GS_GTensor tGS_gRow;                                                         // (CPY,CPY_M,CPY_N)
+    GS_STensor tGS_sRow;                                                         // (CPY,CPY_M,CPY_N)
+    GS_CTensor tGS_cRow;                                                         // (CPY,CPY_M,CPY_N)
+    Tiled_G2S tiled_G2S;
+
+    SR_STensor tSR_sRow;                                                         // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+    SR_RTensor tSR_rRow;                                                         // (CPY,CPY_M,CPY_N,EPI_M,EPI_N) 
+  
+    CTensor tCcRow;                                                              // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+    ThrResidue residue_tCcRow;                                                   // (m, n)
+    ThrNum thr_num;
+    int group;
+    Params const& params;
+
+    CUTLASS_DEVICE void
+    begin() {
+      if (!params.row_broadcast) {
+        fill(tSR_rRow, *(params.ptr_row_array[group]));
+        return;
+      }
+
+      auto synchronize = [&] () { cutlass::arch::NamedBarrier::sync(thr_num, cutlass::arch::ReservedNamedBarriers::EpilogueBarrier); };
+      Tensor tGS_gRow_flt = filter_zeros(tGS_gRow);
+      Tensor tGS_sRow_flt = filter_zeros(tGS_sRow);
+      Tensor tGS_cRow_flt = make_tensor(tGS_cRow.data(), make_layout(tGS_gRow_flt.shape(), tGS_cRow.stride()));
+
+      for (int i = 0; i < size(tGS_gRow_flt); ++i) {
+        if (get<1>(tGS_cRow_flt(i)) >= size<1>(CtaTileShapeMNK{})) {
+          continue; // OOB of SMEM, 
+        }
+        if (elem_less(tGS_cRow_flt(i), make_coord(get<0>(residue_tCcRow), get<1>(residue_tCcRow)))) {
+          tGS_sRow_flt(i) = tGS_gRow_flt(i);
+        }
+        else {
+          tGS_sRow_flt(i) = Element(0); // Set to Zero when OOB so LDS could be issue without any preds.
+        }
+      }
+      synchronize();
+    }
+
+    CUTLASS_DEVICE void
+    begin_loop(int epi_m, int epi_n) {
+      if (epi_m == 0) { // Assumes M-major subtile loop
+        if (!params.row_broadcast) return; // Do not issue LDS when row is scalar 
+        Tensor tSR_sRow_flt = filter_zeros(tSR_sRow(_,_,_,epi_m,epi_n));
+        Tensor tSR_rRow_flt = filter_zeros(tSR_rRow);
+        copy(tSR_sRow_flt, tSR_rRow_flt);
+      }
+    }
+
+    template <typename ElementAccumulator, int FragmentSize>
+    CUTLASS_DEVICE Array<Element, FragmentSize>
+    visit(Array<ElementAccumulator, FragmentSize> const& frg_acc, int epi_v, int epi_m, int epi_n) {
+      Array<Element, FragmentSize> frg_row;
+
+      CUTLASS_PRAGMA_UNROLL
+      for (int i = 0; i < FragmentSize; ++i) {
+        frg_row[i] = tSR_rRow(epi_v * FragmentSize + i);
+      }
+
+      return frg_row;
+    }
+  };
+
+  template <
+    bool ReferenceSrc, // do register tensors reference the src or dst layout of the tiled copy
+    class... Args
+  >
+  CUTLASS_DEVICE auto
+  get_consumer_store_callbacks(ConsumerStoreArgs<Args...> const& args) {
+    auto [M, N, K, L] = args.problem_shape_mnkl;
+    auto [m, n, k, l] = args.tile_coord_mnkl;
+    using ThreadCount = decltype(size(args.tiled_copy));
+
+    Tensor mRow = make_tensor(make_gmem_ptr(params.ptr_row_array[l]), make_shape(M,N,1), params.dRow);
+    Tensor gRow = local_tile(mRow(_,_,l), take<0,2>(args.tile_shape_mnk), make_coord(m, n));          // (CTA_M, CTA_N)
+    Tensor sRow = make_tensor(make_smem_ptr(smem), 
+        make_shape(size<0>(CtaTileShapeMNK{}), size<1>(CtaTileShapeMNK{})), make_shape(_0{}, _1{}));  // (CTA_M, CTA_N)
+    //// G2S: Gmem to Smem
+    auto tiled_g2s = make_tiled_copy(Copy_Atom<DefaultCopy, Element>{},
+                                     Layout< Shape<_1, ThreadCount>, 
+                                            Stride<_0,          _1>>{}, 
+                                     Layout<_1>{});   
+    auto thr_g2s = tiled_g2s.get_slice(args.thread_idx);
+    Tensor tGS_gRow = thr_g2s.partition_S(gRow);
+    Tensor tGS_sRow = thr_g2s.partition_D(sRow);
+
+    //// G2S: Coord 
+    auto cRow = make_identity_tensor(make_shape(size<0>(CtaTileShapeMNK{}), size<1>(CtaTileShapeMNK{})));
+    Tensor tGS_cRow = thr_g2s.partition_S(cRow);
+
+    //// S2R: Smem to Reg
+    Tensor tSR_sRow = sm90_partition_for_epilogue<ReferenceSrc>(sRow, args.epi_tile, args.tiled_copy, args.thread_idx);
+    Tensor tSR_rRow = make_tensor_like(take<0,3>(tSR_sRow));                                           // (CPY,CPY_M,CPY_N)
+
+    return ConsumerStoreCallbacks<decltype(tGS_gRow), decltype(tGS_sRow), decltype(tGS_cRow), decltype(tiled_g2s), decltype(tSR_sRow), decltype(tSR_rRow), decltype(args.tCcD), decltype(args.residue_cD), ThreadCount>(
+      tGS_gRow, 
+      tGS_sRow, 
+      tGS_cRow, tiled_g2s, 
+      tSR_sRow, 
+      tSR_rRow, 
+      args.tCcD, 
+      args.residue_cD,
+      ThreadCount{}, 
+      l,
+      params);
+  }
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+
+// Column vector broadcast
+template<
+  int Stages,
+  class CtaTileShapeMNK,
+  class Element,
+  class StrideMNL = Stride<_1,_0,_0>,
+  int Alignment = 128 / sizeof_bits_v<Element>
+>
+struct Sm90ColOrScalarBroadcastArray {
+  static_assert(Stages == 0, "Column broadcast doesn't support smem usage yet");
+  static_assert(Alignment * sizeof_bits_v<Element> % 128 == 0, "sub-16B alignment not supported yet");
+  static_assert(
+    (cute::is_same_v<StrideMNL, Stride<_1,_0, _0>>) || // col vector broadcast, e.g. per-row alpha/bias
+    (cute::is_same_v<StrideMNL, Stride<_1,_0,int>>));  // batched col vector broadcast, e.g. batched per-row bias
+
+  // Accumulator distributes col elements evenly amongst threads so we can just directly load from gmem
+  struct SharedStorage { };
+
+  // This struct has been modified to have a bool indicating that ptr_col is a 
+  // scalar that must be broadcast, instead of containing a scalar that is 
+  // valid if ptr_col is null.
+  struct Arguments {
+    const Element* const* ptr_col_array = nullptr;
+    bool col_broadcast = true;
+    StrideMNL dCol = {};
+  };
+
+  using Params = Arguments;
+
+  template <class ProblemShape>
+  static constexpr Params
+  to_underlying_arguments(ProblemShape const& problem_shape, Arguments const& args, void* workspace) {
+    return args;
+  }
+
+  template <class ProblemShape>
+  static bool
+  can_implement(ProblemShape const& problem_shape, Arguments const& args) {
+    return true;
+  }
+
+  template <class ProblemShape>
+  static size_t
+  get_workspace_size(ProblemShape const& problem_shape, Arguments const& args) {
+    return 0;
+  }
+
+  template <class ProblemShape>
+  static cutlass::Status
+  initialize_workspace(ProblemShape const& problem_shape, Arguments const& args, void* workspace, cudaStream_t stream,
+    CudaHostAdapter* cuda_adapter = nullptr) {
+    return cutlass::Status::kSuccess;
+  }
+
+  CUTLASS_DEVICE bool
+  is_producer_load_needed() const {
+    return false;
+  }
+
+  CUTLASS_DEVICE bool
+  is_C_load_needed() const {
+    return false;
+  }
+
+  CUTLASS_DEVICE bool
+  is_zero() const {
+    return (!params.col_broadcast && *(params.ptr_col_array[group]) == Element(0));
+  }
+
+  CUTLASS_HOST_DEVICE
+  Sm90ColOrScalarBroadcastArray() { }
+
+  CUTLASS_HOST_DEVICE
+  Sm90ColOrScalarBroadcastArray(Params const& params, SharedStorage const& shared_storage)
+      : params(params) { }
+
+  Params params;
+
+  template <class... Args>
+  CUTLASS_DEVICE auto
+  get_producer_load_callbacks(ProducerLoadArgs<Args...> const& args) {
+    return EmptyProducerLoadCallbacks{};
+  }
+
+  template<class GTensor, class RTensor, class CTensor, class ProblemShape>
+  struct ConsumerStoreCallbacks : EmptyConsumerStoreCallbacks {
+    CUTLASS_DEVICE
+    ConsumerStoreCallbacks(
+      GTensor&& tCgCol,
+      RTensor&& tCrCol,
+      CTensor&& tCcCol,
+      ProblemShape problem_shape,
+      int group,
+      Params const& params
+    ): 
+      tCgCol(cute::forward<GTensor>(tCgCol)),
+      tCrCol(cute::forward<RTensor>(tCrCol)),
+      tCcCol(cute::forward<CTensor>(tCcCol)),
+      m(get<0>(problem_shape)),
+      group(group),
+      params(params) {}
+
+    GTensor tCgCol;                                                                    // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+    RTensor tCrCol;
+    CTensor tCcCol;                                                                    // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+    Params const& params;
+    int m;
+    int group;
+
+    CUTLASS_DEVICE void
+    begin() {
+      Tensor pred = make_tensor<bool>(shape(tCgCol));
+      CUTLASS_PRAGMA_UNROLL
+      for (int i = 0; i < size(pred); ++i) {
+        pred(i) = get<0>(tCcCol(i)) < m;
+      }
+
+      if (!params.col_broadcast) {
+        fill(tCrCol, *(params.ptr_col_array[group]));
+        return;
+      }
+
+      // Filter so we don't issue redundant copies over stride-0 modes
+      // (only works if 0-strides are in same location, which is by construction)
+      copy_if(pred, filter(tCgCol), filter(tCrCol));
+    }
+
+    template <typename ElementAccumulator, int FragmentSize>
+    CUTLASS_DEVICE Array<Element, FragmentSize>
+    visit(Array<ElementAccumulator, FragmentSize> const& frg_acc, int epi_v, int epi_m, int epi_n) {
+      Array<Element, FragmentSize> frg_col;
+      Tensor tCrCol_mn = tCrCol(_,_,_,epi_m,epi_n);
+
+      CUTLASS_PRAGMA_UNROLL
+      for (int i = 0; i < FragmentSize; ++i) {
+        frg_col[i] = tCrCol_mn(epi_v * FragmentSize + i);
+      }
+
+      return frg_col;
+    }
+
+  };
+
+  template <
+    bool ReferenceSrc, // do register tensors reference the src or dst layout of the tiled copy
+    class... Args
+  >
+  CUTLASS_DEVICE auto
+  get_consumer_store_callbacks(ConsumerStoreArgs<Args...> const& args) {
+
+    auto [M, N, K, L] = args.problem_shape_mnkl;
+    auto [m, n, k, l] = args.tile_coord_mnkl;
+
+    Tensor mCol = make_tensor(make_gmem_ptr(params.ptr_col_array[l]), make_shape(M,N,1), params.dCol);
+    Tensor tCgCol = sm90_partition_for_epilogue<ReferenceSrc>(                         // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+      mCol, args.tile_shape_mnk, args.tile_coord_mnkl, args.epi_tile, args.tiled_copy, args.thread_idx);
+    Tensor tCrCol = make_tensor_like(tCgCol);                                          // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+
+    // Generate an identity tensor matching the shape of the global tensor and 
+    //  partition the same way, this will be used to generate the predicate
+    //  tensor for loading
+    Tensor cCol = make_identity_tensor(mCol.shape());
+    Tensor tCcCol = sm90_partition_for_epilogue<ReferenceSrc>(                         // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
+      cCol, args.tile_shape_mnk, args.tile_coord_mnkl, args.epi_tile, args.tiled_copy, args.thread_idx);
+
+    return ConsumerStoreCallbacks(
+      cute::move(tCgCol), 
+      cute::move(tCrCol), 
+      cute::move(tCcCol), 
+      args.problem_shape_mnkl,
+      l,
+      params
+    );
+  }
+};
+
+}

csrc/cutlass_extensions/epilogue/scaled_mm_epilogues_c3x.hpp

@@ -1,6 +1,7 @@
 #pragma once
 
 #include "cutlass_extensions/epilogue/broadcast_load_epilogue_c3x.hpp"
+#include "cutlass_extensions/epilogue/broadcast_load_epilogue_array_c3x.hpp"
 
 /*
    This file defines custom epilogues for fusing channel scales, token scales,
@@ -69,6 +70,16 @@ struct ScaledEpilogueBase {
       0 /*Stages*/, TileShape, T, T, Stride<Int<0>, Int<1>, Int<0>>,
       128 / sizeof_bits_v<T>, EnableNullPtr>;
 
+  template <typename T>
+  using ColOrScalarLoadArray =
+      cutlass::epilogue::fusion::Sm90ColOrScalarBroadcastArray<
+          0 /*Stages*/, TileShape, T, Stride<Int<1>, Int<0>, Int<0>>>;
+
+  template <typename T>
+  using RowOrScalarLoadArray =
+      cutlass::epilogue::fusion::Sm90RowOrScalarBroadcastArray<
+          0 /*Stages*/, TileShape, T, Stride<Int<0>, Int<1>, Int<0>>>;
+
   // This utility function constructs the arguments for the load descriptors
   // from a tensor. It can handle both row and column, as well as row/column or
   // scalar cases.
@@ -96,6 +107,14 @@ struct ScaledEpilogueBase {
                   std::is_same_v<Descriptor, RowLoad<T, true>>);
     return Arguments{data_ptr};
   }
+
+  template <typename Descriptor, typename T>
+  static auto args_from_tensor(const T* const* data_ptr, bool do_broadcast) {
+    using Arguments = typename Descriptor::Arguments;
+    static_assert(std::is_same_v<Descriptor, ColOrScalarLoadArray<T>> ||
+                  std::is_same_v<Descriptor, RowOrScalarLoadArray<T>>);
+    return Arguments{data_ptr, do_broadcast};
+  }
 };
 
 /*
@@ -381,4 +400,51 @@ struct ScaledEpilogueBiasAzpToken
   }
 };
 
+/*
+    This epilogue works like ScaledEpilogue, but ScaleA and ScaleB are pointers
+    to arrays containing different scales used in group gemm. The number of
+   pointers in ScaleA and the number of pointers in ScaleB are equal to the
+   group size.
+*/
+template <typename ElementAcc, typename ElementD, typename EpilogueDescriptor>
+struct ScaledEpilogueArray
+    : private ScaledEpilogueBase<ElementAcc, ElementD, EpilogueDescriptor> {
+ private:
+  using SUPER = ScaledEpilogueBase<ElementAcc, ElementD, EpilogueDescriptor>;
+  using Accum = typename SUPER::Accum;
+  using ScaleA = typename SUPER::template ColOrScalarLoadArray<float>;
+  using ScaleB = typename SUPER::template RowOrScalarLoadArray<float>;
+
+  using Compute0 = cutlass::epilogue::fusion::Sm90Compute<
+      cutlass::multiplies, float, float,
+      cutlass::FloatRoundStyle::round_to_nearest>;
+
+  using EVTCompute0 =
+      cutlass::epilogue::fusion::Sm90EVT<Compute0, ScaleB, Accum>;
+
+  using Compute1 = cutlass::epilogue::fusion::Sm90Compute<
+      cutlass::multiplies, ElementD, float,
+      cutlass::FloatRoundStyle::round_to_nearest>;
+
+ public:
+  using EVTCompute =
+      cutlass::epilogue::fusion::Sm90EVT<Compute1, ScaleA, EVTCompute0>;
+  using ArgumentType = typename EVTCompute::Arguments;
+
+  using ScaleAArray = typename SUPER::template ColOrScalarLoadArray<float>;
+  using ScaleBArray = typename SUPER::template RowOrScalarLoadArray<float>;
+
+  static ArgumentType prepare_args(float const* const* a_scales_ptr,
+                                   float const* const* b_scales_ptr,
+                                   bool a_col_broadcast, bool b_row_broadcast) {
+    auto a_args = SUPER::template args_from_tensor<ScaleAArray, float>(
+        a_scales_ptr, a_col_broadcast);
+    auto b_args = SUPER::template args_from_tensor<ScaleBArray, float>(
+        b_scales_ptr, b_row_broadcast);
+
+    typename EVTCompute0::Arguments evt0_args{b_args, {}, {}};
+    return ArgumentType{a_args, evt0_args, {}};
+  }
+};
+
 };  // namespace vllm::c3x

csrc/ops.h

@@ -164,6 +164,7 @@ int64_t ggml_moe_get_block_size(int64_t type);
 bool cutlass_scaled_mm_supports_fp4(int64_t cuda_device_capability);
 bool cutlass_scaled_mm_supports_fp8(int64_t cuda_device_capability);
 bool cutlass_scaled_mm_supports_block_fp8(int64_t cuda_device_capability);
+bool cutlass_group_gemm_supported(int64_t cuda_device_capability);
 
 void cutlass_scaled_fp4_mm(torch::Tensor& D, torch::Tensor const& A,
                            torch::Tensor const& B, torch::Tensor const& A_sf,
@@ -175,6 +176,19 @@ void cutlass_scaled_mm(torch::Tensor& out, torch::Tensor const& a,
                        torch::Tensor const& b_scales,
                        std::optional<torch::Tensor> const& bias);
 
+void cutlass_moe_mm(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides);
+
+void get_cutlass_moe_mm_data(
+    const torch::Tensor& topk_ids, torch::Tensor& expert_offsets,
+    torch::Tensor& problem_sizes1, torch::Tensor& problem_sizes2,
+    torch::Tensor& input_permutation, torch::Tensor& output_permutation,
+    const int64_t num_experts, const int64_t n, const int64_t k);
+
 void cutlass_scaled_mm_azp(torch::Tensor& out, torch::Tensor const& a,
                            torch::Tensor const& b,
                            torch::Tensor const& a_scales,

csrc/quantization/cutlass_w8a8/moe/get_group_starts.cuh

@@ -0,0 +1,80 @@
+#pragma once
+
+#include <cuda.h>
+#include <torch/all.h>
+#include <c10/cuda/CUDAStream.h>
+
+#include "core/scalar_type.hpp"
+#include "cutlass/bfloat16.h"
+#include "cutlass/float8.h"
+
+template <typename ElementAB, typename ElementC, typename ElementAccumulator>
+__global__ void get_group_gemm_starts(
+    int32_t* expert_offsets, ElementAB** a_offsets, ElementAB** b_offsets,
+    ElementC** out_offsets, ElementAccumulator** a_scales_offsets,
+    ElementAccumulator** b_scales_offsets, ElementAB* a_base_as_int,
+    ElementAB* b_base_as_int, ElementC* out_base_as_int,
+    ElementAccumulator* a_scales_base_as_int,
+    ElementAccumulator* b_scales_base_as_int, int64_t n, int64_t k,
+    bool per_act_token, bool per_out_ch) {
+  int expert_id = threadIdx.x;
+
+  int64_t expert_offset = expert_offsets[expert_id];
+
+  a_offsets[expert_id] = a_base_as_int + expert_offset * k;
+  b_offsets[expert_id] = b_base_as_int + expert_id * k * n;
+  out_offsets[expert_id] = out_base_as_int + expert_offset * n;
+  a_scales_offsets[expert_id] =
+      a_scales_base_as_int + (per_act_token ? expert_offset : 0);
+  b_scales_offsets[expert_id] =
+      b_scales_base_as_int + (per_out_ch ? n * expert_id : expert_id);
+}
+
+#define __CALL_GET_STARTS_KERNEL(TENSOR_C_TYPE, C_TYPE)                    \
+  else if (out_tensors.dtype() == TENSOR_C_TYPE) {                         \
+    get_group_gemm_starts<cutlass::float_e4m3_t, C_TYPE, float>            \
+        <<<1, num_experts, 0, stream>>>(                                   \
+            static_cast<int32_t*>(expert_offsets.data_ptr()),              \
+            static_cast<cutlass::float_e4m3_t**>(a_ptrs.data_ptr()),       \
+            static_cast<cutlass::float_e4m3_t**>(b_ptrs.data_ptr()),       \
+            static_cast<C_TYPE**>(out_ptrs.data_ptr()),                    \
+            static_cast<float**>(a_scales_ptrs.data_ptr()),                \
+            static_cast<float**>(b_scales_ptrs.data_ptr()),                \
+            static_cast<cutlass::float_e4m3_t*>(a_tensors.data_ptr()),     \
+            static_cast<cutlass::float_e4m3_t*>(b_tensors.data_ptr()),     \
+            static_cast<C_TYPE*>(out_tensors.data_ptr()),                  \
+            static_cast<float*>(a_scales.data_ptr()),                      \
+            static_cast<float*>(b_scales.data_ptr()), out_tensors.size(1), \
+            a_tensors.size(1), per_act_token, per_out_ch);                 \
+  }
+
+namespace {
+
+void run_get_group_gemm_starts(
+    torch::Tensor const& expert_offsets, torch::Tensor& a_ptrs,
+    torch::Tensor& b_ptrs, torch::Tensor& out_ptrs,
+    torch::Tensor& a_scales_ptrs, torch::Tensor& b_scales_ptrs,
+    torch::Tensor const& a_tensors, torch::Tensor const& b_tensors,
+    torch::Tensor& out_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales) {
+  TORCH_CHECK(a_tensors.dtype() == torch::kFloat8_e4m3fn);
+  TORCH_CHECK(b_tensors.dtype() == torch::kFloat8_e4m3fn);
+  TORCH_CHECK(a_scales.dtype() == torch::kFloat32);
+  TORCH_CHECK(b_scales.dtype() == torch::kFloat32);
+
+  int num_experts = static_cast<int>(expert_offsets.size(0));
+  bool per_act_token = a_scales.numel() != 1;
+  bool per_out_ch = b_scales.numel() != num_experts;
+
+  auto stream = at::cuda::getCurrentCUDAStream(a_tensors.device().index());
+
+  if (false) {
+  }
+  __CALL_GET_STARTS_KERNEL(torch::kBFloat16, cutlass::bfloat16_t)
+  __CALL_GET_STARTS_KERNEL(torch::kFloat16, half)
+  else {
+    TORCH_CHECK(false, "Invalid output type (must be float16 or bfloat16)");
+  }
+}
+
+}  // namespace

csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu

@@ -0,0 +1,160 @@
+#include <cudaTypedefs.h>
+
+#include <c10/cuda/CUDAGuard.h>
+#include <torch/all.h>
+
+#include "cutlass/cutlass.h"
+#include "grouped_mm_c3x.cuh"
+
+using namespace cute;
+
+namespace {
+
+template <typename InType, typename OutType,
+          template <typename, typename, typename> typename Epilogue>
+struct sm90_fp8_config_default {
+  // M in (16, inf)
+  static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
+  using KernelSchedule =
+      cutlass::gemm::KernelPtrArrayTmaWarpSpecializedPingpongFP8FastAccum;
+  using EpilogueSchedule =
+      cutlass::epilogue::PtrArrayTmaWarpSpecializedPingpong;
+  using TileShape = cute::Shape<cute::_64, cute::_256, cute::_128>;
+  using ClusterShape = cute::Shape<cute::_1, cute::_2, cute::_1>;
+
+  using Cutlass3xGemm =
+      cutlass_3x_group_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
+                            KernelSchedule, EpilogueSchedule>;
+};
+
+template <typename InType, typename OutType,
+          template <typename, typename, typename> typename Epilogue>
+struct sm90_fp8_config_M16 {
+  // M in [1, 16]
+  static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
+  using KernelSchedule =
+      cutlass::gemm::KernelPtrArrayTmaWarpSpecializedPingpongFP8FastAccum;
+  using EpilogueSchedule =
+      cutlass::epilogue::PtrArrayTmaWarpSpecializedPingpong;
+  using TileShape = cute::Shape<cute::_64, cute::_64, cute::_128>;
+  using ClusterShape = cute::Shape<cute::_1, cute::_4, cute::_1>;
+
+  using Cutlass3xGemm =
+      cutlass_3x_group_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
+                            KernelSchedule, EpilogueSchedule>;
+};
+
+template <typename InType, typename OutType,
+          template <typename, typename, typename> typename Epilogue>
+struct sm90_fp8_config_K8192 {
+  // K in [8192, inf)
+  static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
+  using KernelSchedule =
+      cutlass::gemm::KernelPtrArrayTmaWarpSpecializedPingpongFP8FastAccum;
+  using EpilogueSchedule =
+      cutlass::epilogue::PtrArrayTmaWarpSpecializedPingpong;
+  using TileShape = cute::Shape<cute::_128, cute::_128, cute::_128>;
+  using ClusterShape = cute::Shape<cute::_1, cute::_8, cute::_1>;
+
+  using Cutlass3xGemm =
+      cutlass_3x_group_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
+                            KernelSchedule, EpilogueSchedule>;
+};
+
+template <typename InType, typename OutType,
+          template <typename, typename, typename> typename Epilogue>
+struct sm90_fp8_config_N8192 {
+  // N in [8192, inf)
+  static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
+  using KernelSchedule =
+      cutlass::gemm::KernelPtrArrayTmaWarpSpecializedPingpongFP8FastAccum;
+  using EpilogueSchedule =
+      cutlass::epilogue::PtrArrayTmaWarpSpecializedPingpong;
+  using TileShape = cute::Shape<cute::_64, cute::_128, cute::_256>;
+  using ClusterShape = cute::Shape<cute::_1, cute::_8, cute::_1>;
+
+  using Cutlass3xGemm =
+      cutlass_3x_group_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
+                            KernelSchedule, EpilogueSchedule>;
+};
+
+template <typename InType, typename OutType>
+void run_cutlass_moe_mm_sm90(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+  TORCH_CHECK(a_tensors.size(0) > 0, "No input A tensors provided.");
+  TORCH_CHECK(b_tensors.size(0) > 0, "No input B tensors provided.");
+  TORCH_CHECK(out_tensors.size(0) > 0, "No output tensors provided.");
+
+  TORCH_CHECK(a_tensors.dtype() == torch::kFloat8_e4m3fn,
+              "A tensors must be of type float8_e4m3fn.");
+  TORCH_CHECK(b_tensors.dtype() == torch::kFloat8_e4m3fn,
+              "B tensors must be of type float8_e4m3fn.");
+
+  TORCH_CHECK(a_tensors.dtype() == torch::kFloat8_e4m3fn);
+  TORCH_CHECK(b_tensors.dtype() == torch::kFloat8_e4m3fn);
+
+  using Cutlass3xGemmN8192 = typename sm90_fp8_config_N8192<
+      InType, OutType, vllm::c3x::ScaledEpilogueArray>::Cutlass3xGemm;
+  using Cutlass3xGemmK8192 = typename sm90_fp8_config_K8192<
+      InType, OutType, vllm::c3x::ScaledEpilogueArray>::Cutlass3xGemm;
+  using Cutlass3xGemmM16 = typename sm90_fp8_config_M16<
+      InType, OutType, vllm::c3x::ScaledEpilogueArray>::Cutlass3xGemm;
+  using Cutlass3xGemmDefault = typename sm90_fp8_config_default<
+      InType, OutType, vllm::c3x::ScaledEpilogueArray>::Cutlass3xGemm;
+
+  uint32_t const m = a_tensors.size(0);
+  uint32_t const n = out_tensors.size(1);
+  uint32_t const k = a_tensors.size(1);
+
+  if (n >= 8192) {
+    cutlass_group_gemm_caller<Cutlass3xGemmN8192>(
+        out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
+        problem_sizes, a_strides, b_strides, c_strides);
+  } else if (k >= 8192) {
+    cutlass_group_gemm_caller<Cutlass3xGemmK8192>(
+        out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
+        problem_sizes, a_strides, b_strides, c_strides);
+  } else if (m <= 16) {
+    cutlass_group_gemm_caller<Cutlass3xGemmM16>(
+        out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
+        problem_sizes, a_strides, b_strides, c_strides);
+  } else {
+    cutlass_group_gemm_caller<Cutlass3xGemmDefault>(
+        out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
+        problem_sizes, a_strides, b_strides, c_strides);
+  }
+}
+
+void dispatch_moe_mm_sm90(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+  if (out_tensors.dtype() == torch::kBFloat16) {
+    run_cutlass_moe_mm_sm90<cutlass::float_e4m3_t, cutlass::bfloat16_t>(
+        out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
+        problem_sizes, a_strides, b_strides, c_strides);
+  } else {
+    run_cutlass_moe_mm_sm90<cutlass::float_e4m3_t, cutlass::half_t>(
+        out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
+        problem_sizes, a_strides, b_strides, c_strides);
+  }
+}
+
+}  // namespace
+
+void cutlass_moe_mm_sm90(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+  dispatch_moe_mm_sm90(out_tensors, a_tensors, b_tensors, a_scales, b_scales,
+                       expert_offsets, problem_sizes, a_strides, b_strides,
+                       c_strides);
+}

csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cuh

@@ -0,0 +1,149 @@
+#pragma once
+
+#include "cutlass/cutlass.h"
+
+#include "cutlass/gemm/collective/collective_builder.hpp"
+#include "cutlass/epilogue/collective/collective_builder.hpp"
+#include "cutlass/gemm/device/gemm_universal_adapter.h"
+
+#include "cutlass_extensions/epilogue/scaled_mm_epilogues_c3x.hpp"
+#include "cutlass_extensions/common.hpp"
+#include "get_group_starts.cuh"
+
+using namespace cute;
+
+namespace {
+
+using ProblemShape =
+    cutlass::gemm::GroupProblemShape<cute::Shape<int, int, int>>;
+
+using ElementAccumulator = float;
+using ArchTag = cutlass::arch::Sm90;
+using OperatorClass = cutlass::arch::OpClassTensorOp;
+
+using LayoutA = cutlass::layout::RowMajor;
+using LayoutB = cutlass::layout::ColumnMajor;
+using LayoutC = cutlass::layout::RowMajor;
+
+template <typename ElementAB_, typename ElementC_,
+          template <typename, typename, typename> typename Epilogue_,
+          typename TileShape, typename ClusterShape, typename KernelSchedule,
+          typename EpilogueSchedule>
+struct cutlass_3x_group_gemm {
+  using ElementAB = ElementAB_;
+  using ElementC = void;
+  using ElementD = ElementC_;
+  using ElementAccumulator = float;
+
+  using Epilogue = Epilogue_<ElementAccumulator, ElementD, TileShape>;
+
+  using StrideC =
+      cute::remove_pointer_t<cute::Stride<int64_t, cute::Int<1>, cute::Int<0>>>;
+
+  static constexpr int AlignmentAB =
+      128 / cutlass::sizeof_bits<ElementAB>::value;
+  static constexpr int AlignmentC = 128 / cutlass::sizeof_bits<ElementD>::value;
+
+  using EVTCompute = typename Epilogue::EVTCompute;
+
+  using CollectiveEpilogue =
+      typename cutlass::epilogue::collective::CollectiveBuilder<
+          ArchTag, OperatorClass, TileShape, ClusterShape,
+          cutlass::epilogue::collective::EpilogueTileAuto, ElementAccumulator,
+          ElementAccumulator, ElementC, LayoutC*, AlignmentC, ElementD,
+          LayoutC*, AlignmentC, EpilogueSchedule, EVTCompute>::CollectiveOp;
+
+  static constexpr size_t CEStorageSize =
+      sizeof(typename CollectiveEpilogue::SharedStorage);
+  using Stages = typename cutlass::gemm::collective::StageCountAutoCarveout<
+      static_cast<int>(CEStorageSize)>;
+
+  using CollectiveMainloop =
+      typename cutlass::gemm::collective::CollectiveBuilder<
+          ArchTag, OperatorClass, ElementAB, LayoutA*, AlignmentAB, ElementAB,
+          LayoutB*, AlignmentAB, ElementAccumulator, TileShape, ClusterShape,
+          Stages, KernelSchedule>::CollectiveOp;
+
+  using KernelType = enable_sm90_only<cutlass::gemm::kernel::GemmUniversal<
+      ProblemShape, CollectiveMainloop, CollectiveEpilogue>>;
+
+  struct GemmKernel : public KernelType {};
+};
+
+template <typename Gemm>
+void cutlass_group_gemm_caller(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+  using ElementAB = typename Gemm::ElementAB;
+  using ElementD = typename Gemm::ElementD;
+
+  int num_experts = static_cast<int>(expert_offsets.size(0));
+  int k_size = a_tensors.size(1);
+  int n_size = out_tensors.size(1);
+
+  bool per_act_token = a_scales.numel() != 1;
+  bool per_out_ch = b_scales.numel() != num_experts;
+
+  auto stream = at::cuda::getCurrentCUDAStream(a_tensors.device().index());
+
+  auto options_int =
+      torch::TensorOptions().dtype(torch::kInt64).device(a_tensors.device());
+
+  torch::Tensor a_ptrs = torch::empty(num_experts, options_int);
+  torch::Tensor b_ptrs = torch::empty(num_experts, options_int);
+  torch::Tensor out_ptrs = torch::empty(num_experts, options_int);
+  torch::Tensor a_scales_ptrs = torch::empty(num_experts, options_int);
+  torch::Tensor b_scales_ptrs = torch::empty(num_experts, options_int);
+
+  run_get_group_gemm_starts(expert_offsets, a_ptrs, b_ptrs, out_ptrs,
+                            a_scales_ptrs, b_scales_ptrs, a_tensors, b_tensors,
+                            out_tensors, a_scales, b_scales);
+
+  using GemmKernel = typename Gemm::GemmKernel;
+  using StrideA = Stride<int64_t, Int<1>, Int<0>>;
+  using StrideB = Stride<int64_t, Int<1>, Int<0>>;
+  using StrideC = typename GemmKernel::InternalStrideC;
+
+  ProblemShape::UnderlyingProblemShape* problem_sizes_as_shapes =
+      static_cast<ProblemShape::UnderlyingProblemShape*>(
+          problem_sizes.data_ptr());
+  ProblemShape prob_shape{num_experts, problem_sizes_as_shapes, nullptr};
+
+  typename GemmKernel::MainloopArguments mainloop_args{
+      static_cast<const ElementAB**>(a_ptrs.data_ptr()),
+      static_cast<StrideA*>(a_strides.data_ptr()),
+      static_cast<const ElementAB**>(b_ptrs.data_ptr()),
+      static_cast<StrideB*>(b_strides.data_ptr())};
+
+  // Currently, we are only able to do broadcast on either all or none a_scales
+  // and on either all or none b_scales
+  typename GemmKernel::EpilogueArguments epilogue_args{
+      Gemm::Epilogue::prepare_args(
+          static_cast<const ElementAccumulator**>(a_scales_ptrs.data_ptr()),
+          static_cast<const ElementAccumulator**>(b_scales_ptrs.data_ptr()),
+          per_act_token, per_out_ch),
+      nullptr, static_cast<StrideC*>(c_strides.data_ptr()),
+      static_cast<ElementD**>(out_ptrs.data_ptr()),
+      static_cast<StrideC*>(c_strides.data_ptr())};
+
+  typename GemmKernel::Arguments args{
+      cutlass::gemm::GemmUniversalMode::kGrouped, prob_shape, mainloop_args,
+      epilogue_args};
+
+  using GemmOp = cutlass::gemm::device::GemmUniversalAdapter<GemmKernel>;
+  GemmOp gemm_op;
+  CUTLASS_CHECK(gemm_op.can_implement(args));
+
+  size_t workspace_size = gemm_op.get_workspace_size(args);
+  auto const workspace_options =
+      torch::TensorOptions().dtype(torch::kUInt8).device(a_tensors.device());
+  auto workspace = torch::empty(workspace_size, workspace_options);
+
+  cutlass::Status status = gemm_op.run(args, workspace.data_ptr(), stream);
+  CUTLASS_CHECK(status);
+}
+
+}  // namespace

csrc/quantization/cutlass_w8a8/moe/moe_data.cu

@@ -0,0 +1,90 @@
+#include <cudaTypedefs.h>
+
+#include <c10/cuda/CUDAGuard.h>
+#include <torch/all.h>
+
+#include <iostream>
+
+constexpr uint64_t THREADS_PER_EXPERT = 512;
+
+__global__ void compute_problem_sizes(const int* __restrict__ topk_ids,
+                                      int32_t* problem_sizes1,
+                                      int32_t* problem_sizes2,
+                                      int32_t* atomic_buffer,
+                                      const int topk_length, const int n,
+                                      const int k) {
+  int expert_id = blockIdx.x;
+
+  int occurrences = 0;
+  for (int i = threadIdx.x; i < topk_length; i += THREADS_PER_EXPERT) {
+    occurrences += (topk_ids[i] == expert_id);
+  }
+  atomicAdd(&atomic_buffer[expert_id], occurrences);
+  __syncthreads();
+
+  if (threadIdx.x == 0) {
+    int final_occurrences = atomic_buffer[expert_id];
+    problem_sizes1[expert_id * 3] = final_occurrences;
+    problem_sizes1[expert_id * 3 + 1] = 2 * n;
+    problem_sizes1[expert_id * 3 + 2] = k;
+    problem_sizes2[expert_id * 3] = final_occurrences;
+    problem_sizes2[expert_id * 3 + 1] = k;
+    problem_sizes2[expert_id * 3 + 2] = n;
+  }
+}
+
+__global__ void compute_expert_offsets(
+    const int32_t* __restrict__ problem_sizes1, int32_t* expert_offsets,
+    int32_t* atomic_buffer, const int num_experts) {
+  int32_t tot_offset = 0;
+  expert_offsets[0] = 0;
+  for (int i = 0; i < num_experts; ++i) {
+    atomic_buffer[i] = tot_offset;
+    tot_offset += problem_sizes1[i * 3];
+    expert_offsets[i + 1] = tot_offset;
+  }
+}
+
+__global__ void compute_arg_sorts(const int* __restrict__ topk_ids,
+                                  int32_t* input_permutation,
+                                  int32_t* output_permutation,
+                                  int32_t* atomic_buffer, const int topk_length,
+                                  const int topk) {
+  int expert_id = blockIdx.x;
+
+  for (int i = threadIdx.x; i < topk_length; i += THREADS_PER_EXPERT) {
+    if (topk_ids[i] == expert_id) {
+      int start = atomicAdd(&atomic_buffer[expert_id], 1);
+      input_permutation[start] = i / topk;
+      output_permutation[i] = start;
+    }
+  }
+}
+
+void get_cutlass_moe_mm_data_caller(
+    const torch::Tensor& topk_ids, torch::Tensor& expert_offsets,
+    torch::Tensor& problem_sizes1, torch::Tensor& problem_sizes2,
+    torch::Tensor& input_permutation, torch::Tensor& output_permutation,
+    const int64_t num_experts, const int64_t n, const int64_t k) {
+  auto stream = at::cuda::getCurrentCUDAStream(topk_ids.device().index());
+  auto options_int32 =
+      torch::TensorOptions().dtype(torch::kInt32).device(topk_ids.device());
+  torch::Tensor atomic_buffer = torch::zeros(num_experts, options_int32);
+
+  int num_threads = min(THREADS_PER_EXPERT, topk_ids.numel());
+  compute_problem_sizes<<<num_experts, num_threads, 0, stream>>>(
+      static_cast<const int32_t*>(topk_ids.data_ptr()),
+      static_cast<int32_t*>(problem_sizes1.data_ptr()),
+      static_cast<int32_t*>(problem_sizes2.data_ptr()),
+      static_cast<int32_t*>(atomic_buffer.data_ptr()), topk_ids.numel(), n, k);
+  compute_expert_offsets<<<1, 1, 0, stream>>>(
+      static_cast<const int32_t*>(problem_sizes1.data_ptr()),
+      static_cast<int32_t*>(expert_offsets.data_ptr()),
+      static_cast<int32_t*>(atomic_buffer.data_ptr()), num_experts);
+  compute_arg_sorts<<<num_experts, num_threads, 0, stream>>>(
+      static_cast<const int32_t*>(topk_ids.data_ptr()),
+      static_cast<int32_t*>(input_permutation.data_ptr()),
+      static_cast<int32_t*>(output_permutation.data_ptr()),
+      static_cast<int32_t*>(atomic_buffer.data_ptr()), topk_ids.numel(),
+      topk_ids.size(1));
+}

csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu

@@ -29,6 +29,20 @@ void cutlass_scaled_mm_sm90(torch::Tensor& c, torch::Tensor const& a,
                             torch::Tensor const& a_scales,
                             torch::Tensor const& b_scales,
                             std::optional<torch::Tensor> const& bias);
+
+void cutlass_moe_mm_sm90(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides);
+
+void get_cutlass_moe_mm_data_caller(
+    const torch::Tensor& topk_ids, torch::Tensor& expert_offsets,
+    torch::Tensor& problem_sizes1, torch::Tensor& problem_sizes2,
+    torch::Tensor& input_permutation, torch::Tensor& output_permutation,
+    const int64_t num_experts, const int64_t n, const int64_t k);
+
 #endif
 
 #if defined ENABLE_SCALED_MM_SM100 && ENABLE_SCALED_MM_SM100
@@ -102,6 +116,19 @@ bool cutlass_scaled_mm_supports_block_fp8(int64_t cuda_device_capability) {
   return false;
 }
 
+bool cutlass_group_gemm_supported(int64_t cuda_device_capability) {
+  // CUTLASS groped FP8 kernels need at least CUDA 12.3
+  // and SM90 (Hopper)
+
+#if defined CUDA_VERSION
+  if (cuda_device_capability == 90) {
+    return CUDA_VERSION >= 12030;
+  }
+#endif
+
+  return false;
+}
+
 void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a,
                        torch::Tensor const& b, torch::Tensor const& a_scales,
                        torch::Tensor const& b_scales,
@@ -168,6 +195,46 @@ void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a,
       version_num);
 }
 
+void cutlass_moe_mm(
+    torch::Tensor& out_tensors, torch::Tensor const& a_tensors,
+    torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
+    torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
+    torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+  int32_t version_num = get_sm_version_num();
+#if defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90
+  cutlass_moe_mm_sm90(out_tensors, a_tensors, b_tensors, a_scales, b_scales,
+                      expert_offsets, problem_sizes, a_strides, b_strides,
+                      c_strides);
+  return;
+#endif
+  TORCH_CHECK_NOT_IMPLEMENTED(
+      false,
+      "No compiled cutlass_scaled_mm for CUDA device capability: ", version_num,
+      ". Required capability: 90");
+}
+
+void get_cutlass_moe_mm_data(
+    const torch::Tensor& topk_ids, torch::Tensor& expert_offsets,
+    torch::Tensor& problem_sizes1, torch::Tensor& problem_sizes2,
+    torch::Tensor& input_permutation, torch::Tensor& output_permutation,
+    const int64_t num_experts, const int64_t n, const int64_t k) {
+  // This function currently gets compiled only if we have a valid cutlass moe
+  // mm to run it for.
+  int32_t version_num = get_sm_version_num();
+#if defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90
+  get_cutlass_moe_mm_data_caller(topk_ids, expert_offsets, problem_sizes1,
+                                 problem_sizes2, input_permutation,
+                                 output_permutation, num_experts, n, k);
+  return;
+#endif
+  TORCH_CHECK_NOT_IMPLEMENTED(
+      false,
+      "No compiled get_cutlass_moe_mm_data: no cutlass_scaled_mm kernel for "
+      "CUDA device capability: ",
+      version_num, ". Required capability: 90");
+}
+
 void cutlass_scaled_mm_azp(torch::Tensor& c, torch::Tensor const& a,
                            torch::Tensor const& b,
                            torch::Tensor const& a_scales,

csrc/quantization/gguf/gguf_kernel.cu

@@ -375,25 +375,25 @@ torch::Tensor ggml_moe_a8(torch::Tensor X,  // input
 int64_t ggml_moe_get_block_size(int64_t type) {
   switch (type) {
     case 2:
-      return MMQ_X_Q4_0;
+      return MOE_X_Q4_0;
     case 3:
-      return MMQ_X_Q4_1;
+      return MOE_X_Q4_1;
     case 6:
-      return MMQ_X_Q5_0;
+      return MOE_X_Q5_0;
     case 7:
-      return MMQ_X_Q5_1;
+      return MOE_X_Q5_1;
     case 8:
-      return MMQ_X_Q8_0;
+      return MOE_X_Q8_0;
     case 10:
-      return MMQ_X_Q2_K;
+      return MOE_X_Q2_K;
     case 11:
-      return MMQ_X_Q3_K;
+      return MOE_X_Q3_K;
     case 12:
-      return MMQ_X_Q4_K;
+      return MOE_X_Q4_K;
     case 13:
-      return MMQ_X_Q5_K;
+      return MOE_X_Q5_K;
     case 14:
-      return MMQ_X_Q6_K;
+      return MOE_X_Q6_K;
   }
   return 0;
 }

csrc/quantization/gguf/moe.cuh

@@ -129,12 +129,12 @@ static __device__ __forceinline__ void moe_q(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q4_0 64
-  #define MMQ_Y_Q4_0 128
+  #define MOE_X_Q4_0 64
+  #define MOE_Y_Q4_0 128
   #define NWARPS_Q4_0 8
 #else
-  #define MMQ_X_Q4_0 4
-  #define MMQ_Y_Q4_0 32
+  #define MOE_X_Q4_0 4
+  #define MOE_Y_Q4_0 32
   #define NWARPS_Q4_0 4
 #endif
 
@@ -149,8 +149,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q4_0, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q4_0;
-  const int mmq_y = MMQ_Y_Q4_0;
+  const int mmq_x = MOE_X_Q4_0;
+  const int mmq_y = MOE_Y_Q4_0;
   const int nwarps = NWARPS_Q4_0;
 
   moe_q<scalar_t, QK4_0, QR4_0, QI4_0, true, block_q4_0, mmq_x, mmq_y, nwarps,
@@ -167,8 +167,8 @@ static void ggml_moe_q4_0_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  int mmq_x = MMQ_X_Q4_0;
-  int mmq_y = MMQ_Y_Q4_0;
+  int mmq_x = MOE_X_Q4_0;
+  int mmq_y = MOE_Y_Q4_0;
   int nwarps = NWARPS_Q4_0;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -190,12 +190,12 @@ static void ggml_moe_q4_0_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q4_1 64
-  #define MMQ_Y_Q4_1 128
+  #define MOE_X_Q4_1 64
+  #define MOE_Y_Q4_1 128
   #define NWARPS_Q4_1 8
 #else
-  #define MMQ_X_Q4_1 4
-  #define MMQ_Y_Q4_1 32
+  #define MOE_X_Q4_1 4
+  #define MOE_Y_Q4_1 32
   #define NWARPS_Q4_1 4
 #endif
 
@@ -210,8 +210,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q4_1, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q4_1;
-  const int mmq_y = MMQ_Y_Q4_1;
+  const int mmq_x = MOE_X_Q4_1;
+  const int mmq_y = MOE_Y_Q4_1;
   const int nwarps = NWARPS_Q4_1;
 
   moe_q<scalar_t, QK4_1, QR4_1, QI4_1, true, block_q4_1, mmq_x, mmq_y, nwarps,
@@ -228,8 +228,8 @@ static void ggml_moe_q4_1_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  int mmq_x = MMQ_X_Q4_1;
-  int mmq_y = MMQ_Y_Q4_1;
+  int mmq_x = MOE_X_Q4_1;
+  int mmq_y = MOE_Y_Q4_1;
   int nwarps = NWARPS_Q4_1;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -251,12 +251,12 @@ static void ggml_moe_q4_1_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q5_0 64
-  #define MMQ_Y_Q5_0 128
+  #define MOE_X_Q5_0 64
+  #define MOE_Y_Q5_0 128
   #define NWARPS_Q5_0 8
 #else
-  #define MMQ_X_Q5_0 4
-  #define MMQ_Y_Q5_0 32
+  #define MOE_X_Q5_0 4
+  #define MOE_Y_Q5_0 32
   #define NWARPS_Q5_0 4
 #endif
 
@@ -271,8 +271,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q5_0, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q5_0;
-  const int mmq_y = MMQ_Y_Q5_0;
+  const int mmq_x = MOE_X_Q5_0;
+  const int mmq_y = MOE_Y_Q5_0;
   const int nwarps = NWARPS_Q5_0;
 
   moe_q<scalar_t, QK5_0, QR5_0, QI5_0, false, block_q5_0, mmq_x, mmq_y, nwarps,
@@ -289,8 +289,8 @@ static void ggml_moe_q5_0_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q5_0;
-  const int mmq_y = MMQ_Y_Q5_0;
+  const int mmq_x = MOE_X_Q5_0;
+  const int mmq_y = MOE_Y_Q5_0;
   const int nwarps = NWARPS_Q5_0;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -312,12 +312,12 @@ static void ggml_moe_q5_0_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q5_1 64
-  #define MMQ_Y_Q5_1 128
+  #define MOE_X_Q5_1 64
+  #define MOE_Y_Q5_1 128
   #define NWARPS_Q5_1 8
 #else
-  #define MMQ_X_Q5_1 4
-  #define MMQ_Y_Q5_1 32
+  #define MOE_X_Q5_1 4
+  #define MOE_Y_Q5_1 32
   #define NWARPS_Q5_1 4
 #endif
 
@@ -332,8 +332,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q5_1, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q5_1;
-  const int mmq_y = MMQ_Y_Q5_1;
+  const int mmq_x = MOE_X_Q5_1;
+  const int mmq_y = MOE_Y_Q5_1;
   const int nwarps = NWARPS_Q5_1;
 
   moe_q<scalar_t, QK5_1, QR5_1, QI5_1, true, block_q5_1, mmq_x, mmq_y, nwarps,
@@ -350,8 +350,8 @@ static void ggml_moe_q5_1_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q5_1;
-  const int mmq_y = MMQ_Y_Q5_1;
+  const int mmq_x = MOE_X_Q5_1;
+  const int mmq_y = MOE_Y_Q5_1;
   const int nwarps = NWARPS_Q5_1;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -373,12 +373,12 @@ static void ggml_moe_q5_1_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q8_0 64
-  #define MMQ_Y_Q8_0 128
+  #define MOE_X_Q8_0 64
+  #define MOE_Y_Q8_0 128
   #define NWARPS_Q8_0 8
 #else
-  #define MMQ_X_Q8_0 4
-  #define MMQ_Y_Q8_0 32
+  #define MOE_X_Q8_0 4
+  #define MOE_Y_Q8_0 32
   #define NWARPS_Q8_0 4
 #endif
 
@@ -393,8 +393,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q8_0, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q8_0;
-  const int mmq_y = MMQ_Y_Q8_0;
+  const int mmq_x = MOE_X_Q8_0;
+  const int mmq_y = MOE_Y_Q8_0;
   const int nwarps = NWARPS_Q8_0;
 
   moe_q<scalar_t, QK8_0, QR8_0, QI8_0, false, block_q8_0, mmq_x, mmq_y, nwarps,
@@ -411,8 +411,8 @@ static void ggml_moe_q8_0_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q8_0;
-  const int mmq_y = MMQ_Y_Q8_0;
+  const int mmq_x = MOE_X_Q8_0;
+  const int mmq_y = MOE_Y_Q8_0;
   const int nwarps = NWARPS_Q8_0;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -434,12 +434,12 @@ static void ggml_moe_q8_0_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q2_K 64
-  #define MMQ_Y_Q2_K 128
+  #define MOE_X_Q2_K 64
+  #define MOE_Y_Q2_K 128
   #define NWARPS_Q2_K 8
 #else
-  #define MMQ_X_Q2_K 4
-  #define MMQ_Y_Q2_K 32
+  #define MOE_X_Q2_K 4
+  #define MOE_Y_Q2_K 32
   #define NWARPS_Q2_K 4
 #endif
 
@@ -454,8 +454,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q2_K, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q2_K;
-  const int mmq_y = MMQ_Y_Q2_K;
+  const int mmq_x = MOE_X_Q2_K;
+  const int mmq_y = MOE_Y_Q2_K;
   const int nwarps = NWARPS_Q2_K;
 
   moe_q<scalar_t, QK_K, QR2_K, QI2_K, false, block_q2_K, mmq_x, mmq_y, nwarps,
@@ -472,8 +472,8 @@ static void ggml_moe_q2_K_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q2_K;
-  const int mmq_y = MMQ_Y_Q2_K;
+  const int mmq_x = MOE_X_Q2_K;
+  const int mmq_y = MOE_Y_Q2_K;
   const int nwarps = NWARPS_Q2_K;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -495,12 +495,12 @@ static void ggml_moe_q2_K_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q3_K 64
-  #define MMQ_Y_Q3_K 128
+  #define MOE_X_Q3_K 64
+  #define MOE_Y_Q3_K 128
   #define NWARPS_Q3_K 8
 #else
-  #define MMQ_X_Q3_K 4
-  #define MMQ_Y_Q3_K 32
+  #define MOE_X_Q3_K 4
+  #define MOE_Y_Q3_K 32
   #define NWARPS_Q3_K 4
 #endif
 
@@ -516,8 +516,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q3_K, 2)
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
 
-  const int mmq_x = MMQ_X_Q3_K;
-  const int mmq_y = MMQ_Y_Q3_K;
+  const int mmq_x = MOE_X_Q3_K;
+  const int mmq_y = MOE_Y_Q3_K;
   const int nwarps = NWARPS_Q3_K;
 
   moe_q<scalar_t, QK_K, QR3_K, QI3_K, false, block_q3_K, mmq_x, mmq_y, nwarps,
@@ -533,8 +533,8 @@ static void ggml_moe_q3_K_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q3_K;
-  const int mmq_y = MMQ_Y_Q3_K;
+  const int mmq_x = MOE_X_Q3_K;
+  const int mmq_y = MOE_Y_Q3_K;
   const int nwarps = NWARPS_Q3_K;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -556,12 +556,12 @@ static void ggml_moe_q3_K_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q4_K 64
-  #define MMQ_Y_Q4_K 128
+  #define MOE_X_Q4_K 64
+  #define MOE_Y_Q4_K 128
   #define NWARPS_Q4_K 8
 #else
-  #define MMQ_X_Q4_K 4
-  #define MMQ_Y_Q4_K 32
+  #define MOE_X_Q4_K 4
+  #define MOE_Y_Q4_K 32
   #define NWARPS_Q4_K 4
 #endif
 
@@ -576,8 +576,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q4_K, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q4_K;
-  const int mmq_y = MMQ_Y_Q4_K;
+  const int mmq_x = MOE_X_Q4_K;
+  const int mmq_y = MOE_Y_Q4_K;
   const int nwarps = NWARPS_Q4_K;
 
   moe_q<scalar_t, QK_K, QR4_K, QI4_K, true, block_q4_K, mmq_x, mmq_y, nwarps,
@@ -594,8 +594,8 @@ static void ggml_moe_q4_K_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q4_K;
-  const int mmq_y = MMQ_Y_Q4_K;
+  const int mmq_x = MOE_X_Q4_K;
+  const int mmq_y = MOE_Y_Q4_K;
   const int nwarps = NWARPS_Q4_K;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -617,12 +617,12 @@ static void ggml_moe_q4_K_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q5_K 64
-  #define MMQ_Y_Q5_K 128
+  #define MOE_X_Q5_K 64
+  #define MOE_Y_Q5_K 128
   #define NWARPS_Q5_K 8
 #else
-  #define MMQ_X_Q5_K 4
-  #define MMQ_Y_Q5_K 32
+  #define MOE_X_Q5_K 4
+  #define MOE_Y_Q5_K 32
   #define NWARPS_Q5_K 4
 #endif
 
@@ -637,8 +637,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q5_K, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q5_K;
-  const int mmq_y = MMQ_Y_Q5_K;
+  const int mmq_x = MOE_X_Q5_K;
+  const int mmq_y = MOE_Y_Q5_K;
   const int nwarps = NWARPS_Q5_K;
 
   moe_q<scalar_t, QK_K, QR5_K, QI5_K, true, block_q5_K, mmq_x, mmq_y, nwarps,
@@ -655,8 +655,8 @@ static void ggml_moe_q5_K_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q5_K;
-  const int mmq_y = MMQ_Y_Q5_K;
+  const int mmq_x = MOE_X_Q5_K;
+  const int mmq_y = MOE_Y_Q5_K;
   const int nwarps = NWARPS_Q5_K;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;
@@ -678,12 +678,12 @@ static void ggml_moe_q5_K_q8_1_cuda(
 }
 
 #if defined(USE_ROCM)
-  #define MMQ_X_Q6_K 64
-  #define MMQ_Y_Q6_K 128
+  #define MOE_X_Q6_K 64
+  #define MOE_Y_Q6_K 128
   #define NWARPS_Q6_K 8
 #else
-  #define MMQ_X_Q6_K 4
-  #define MMQ_Y_Q6_K 32
+  #define MOE_X_Q6_K 4
+  #define MOE_Y_Q6_K 32
   #define NWARPS_Q6_K 4
 #endif
 
@@ -698,8 +698,8 @@ __launch_bounds__(WARP_SIZE_GGUF* NWARPS_Q6_K, 2)
              const int exp_stride, const int ncols_x, const int nrows_x,
              const int ncols_y, const int nrows_y, const int nrows_dst,
              const int top_k) {
-  const int mmq_x = MMQ_X_Q6_K;
-  const int mmq_y = MMQ_Y_Q6_K;
+  const int mmq_x = MOE_X_Q6_K;
+  const int mmq_y = MOE_Y_Q6_K;
   const int nwarps = NWARPS_Q6_K;
 
   moe_q<scalar_t, QK_K, QR6_K, QI6_K, false, block_q6_K, mmq_x, mmq_y, nwarps,
@@ -716,8 +716,8 @@ static void ggml_moe_q6_K_q8_1_cuda(
     const int exp_stride, const int ncols_x, const int nrows_x,
     const int ncols_y, const int nrows_y, const int nrows_dst, const int top_k,
     const int tokens_post_padded, cudaStream_t stream) {
-  const int mmq_x = MMQ_X_Q6_K;
-  const int mmq_y = MMQ_Y_Q6_K;
+  const int mmq_x = MOE_X_Q6_K;
+  const int mmq_y = MOE_Y_Q6_K;
   const int nwarps = NWARPS_Q6_K;
 
   const int block_num_x = (nrows_x + mmq_y - 1) / mmq_y;

csrc/quantization/gptq_marlin/awq_marlin_repack.cu

@@ -14,7 +14,7 @@ __global__ void awq_marlin_repack_kernel(
   int n_tiles = size_n / tile_n_size;
   int block_k_tiles = div_ceil(k_tiles, gridDim.x);
 
-  int start_k_tile = blockIdx.x * block_k_tiles;
+  auto start_k_tile = blockIdx.x * block_k_tiles;
   if (start_k_tile >= k_tiles) {
     return;
   }
@@ -51,8 +51,8 @@ __global__ void awq_marlin_repack_kernel(
     int4* sh_ptr = sh + stage_size * pipe;
 
     if (threadIdx.x < stage_size) {
-      int k_id = threadIdx.x / stage_n_threads;
-      int n_id = threadIdx.x % stage_n_threads;
+      auto k_id = threadIdx.x / stage_n_threads;
+      auto n_id = threadIdx.x % stage_n_threads;
 
       int first_k = k_tile_id * tile_k_size;
 
@@ -70,8 +70,8 @@ __global__ void awq_marlin_repack_kernel(
       return;
     }
 
-    int warp_id = threadIdx.x / 32;
-    int th_id = threadIdx.x % 32;
+    auto warp_id = threadIdx.x / 32;
+    auto th_id = threadIdx.x % 32;
 
     if (warp_id >= 4) {
       return;
@@ -265,4 +265,4 @@ TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, CUDA, m) {
 
 TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, Meta, m) {
   m.impl("awq_marlin_repack", &awq_marlin_repack_meta);
-}
+}

csrc/quantization/gptq_marlin/gptq_marlin.cu

@@ -42,7 +42,7 @@ namespace marlin {
 __global__ void permute_cols_kernel(int4 const* __restrict__ a_int4_ptr,
                                     int const* __restrict__ perm_int_ptr,
                                     int4* __restrict__ out_int4_ptr, int size_m,
-                                    int size_k, int block_rows) {}
+                                    int size_k, int lda, int block_rows) {}
 
 template <typename scalar_t,  // compute dtype, half or nv_float16
           const vllm::ScalarTypeId w_type_id,  // weight ScalarType id
@@ -459,29 +459,32 @@ __device__ inline void barrier_release(int* lock, bool reset = false) {
 __global__ void permute_cols_kernel(int4 const* __restrict__ a_int4_ptr,
                                     int const* __restrict__ perm_int_ptr,
                                     int4* __restrict__ out_int4_ptr, int size_m,
-                                    int size_k, int block_rows) {
-  int start_row = block_rows * blockIdx.x;
+                                    int size_k, int lda, int block_rows) {
+  auto start_row = block_rows * blockIdx.x;
   int finish_row = start_row + block_rows;
   if (finish_row > size_m) {
     finish_row = size_m;
   }
   int cur_block_rows = finish_row - start_row;
 
-  int row_stride = size_k * sizeof(half) / 16;
+  int input_row_stride = lda * sizeof(half) / 16;
+  int output_row_stride = size_k * sizeof(half) / 16;
 
   auto permute_row = [&](int row) {
     int iters = size_k / default_threads;
     int rest = size_k % default_threads;
 
-    int offset = row * row_stride;
+    int input_offset = row * input_row_stride;
+    int output_offset = row * output_row_stride;
 
-    half const* a_row_half = reinterpret_cast<half const*>(a_int4_ptr + offset);
-    half* out_half = reinterpret_cast<half*>(out_int4_ptr + offset);
+    half const* a_row_half =
+        reinterpret_cast<half const*>(a_int4_ptr + input_offset);
+    half* out_half = reinterpret_cast<half*>(out_int4_ptr + output_offset);
 
     int base_k = 0;
 
     for (int i = 0; i < iters; i++) {
-      int cur_k = base_k + threadIdx.x;
+      auto cur_k = base_k + threadIdx.x;
       int src_pos = perm_int_ptr[cur_k];
 
       out_half[cur_k] = a_row_half[src_pos];
@@ -491,7 +494,7 @@ __global__ void permute_cols_kernel(int4 const* __restrict__ a_int4_ptr,
 
     if (rest) {
       if (threadIdx.x < rest) {
-        int cur_k = base_k + threadIdx.x;
+        auto cur_k = base_k + threadIdx.x;
         int src_pos = perm_int_ptr[cur_k];
 
         out_half[cur_k] = a_row_half[src_pos];
@@ -537,6 +540,7 @@ __global__ void Marlin(
     int prob_m,           // batch dimension m
     int prob_n,           // output dimension n
     int prob_k,           // reduction dimension k
+    int lda,              // A.stride(0), equal to prob_k is A is contiguous
     int* locks,           // extra global storage for barrier synchronization
     bool use_atomic_add,  // whether to use atomic add to reduce
     bool use_fp32_reduce  // whether to use fp32 global reduce
@@ -600,7 +604,7 @@ __global__ void Marlin(
   // We can easily implement parallel problem execution by just remapping
   // indices and advancing global pointers
   if (slice_col_par >= n_tiles) {
-    A += (slice_col_par / n_tiles) * 16 * thread_m_blocks * prob_k / 8;
+    A += (slice_col_par / n_tiles) * 16 * thread_m_blocks * lda / 8;
     C += (slice_col_par / n_tiles) * 16 * thread_m_blocks * prob_n / 8;
     locks += (slice_col_par / n_tiles) * n_tiles;
     slice_col = slice_col_par % n_tiles;
@@ -631,7 +635,7 @@ __global__ void Marlin(
       }
     }
     if (slice_col == n_tiles) {
-      A += 16 * thread_m_blocks * prob_k / 8;
+      A += 16 * thread_m_blocks * lda / 8;
       C += 16 * thread_m_blocks * prob_n / 8;
       locks += n_tiles;
       slice_col = 0;
@@ -643,7 +647,7 @@ __global__ void Marlin(
   // A sizes/strides
 
   // stride of the A matrix in global memory
-  int a_gl_stride = prob_k / 8;
+  int a_gl_stride = lda / 8;
   // stride of an A matrix tile in shared memory
   constexpr int a_sh_stride = 16 * thread_k_blocks / 8;
   // delta between subsequent A tiles in global memory
@@ -719,8 +723,8 @@ __global__ void Marlin(
                 (threadIdx.x % b_sh_stride_threads) * b_thread_vecs;
   b_gl_rd += b_sh_stride * slice_col;
   b_gl_rd += b_gl_rd_delta_o * slice_row;
-  int b_sh_wr = threadIdx.x * b_thread_vecs;
-  int b_sh_rd = threadIdx.x * b_thread_vecs;
+  auto b_sh_wr = threadIdx.x * b_thread_vecs;
+  auto b_sh_rd = threadIdx.x * b_thread_vecs;
 
   // For act_order
   constexpr int k_iter_size = tb_k / b_sh_wr_iters;
@@ -739,7 +743,7 @@ __global__ void Marlin(
                 s_sh_stride * slice_col + threadIdx.x;
     }
   }
-  int s_sh_wr = threadIdx.x;
+  auto s_sh_wr = threadIdx.x;
   bool s_sh_wr_pred = threadIdx.x < s_sh_stride;
 
   // Zero-points
@@ -752,7 +756,7 @@ __global__ void Marlin(
                  zp_sh_stride * slice_col + threadIdx.x;
     }
   }
-  int zp_sh_wr = threadIdx.x;
+  auto zp_sh_wr = threadIdx.x;
   bool zp_sh_wr_pred = threadIdx.x < zp_sh_stride;
 
   // We use a different scale layout for grouped and column-wise quantization as
@@ -1043,7 +1047,7 @@ __global__ void Marlin(
           int4* sh_s_stage = sh_s + s_sh_stage * pipe;
           reinterpret_cast<int4*>(&frag_s[k % 2])[0] = sh_s_stage[s_sh_rd];
         } else {
-          int warp_id = threadIdx.x / 32;
+          auto warp_id = threadIdx.x / 32;
           int n_warps = thread_n_blocks / 4;
 
           int warp_row = warp_id / n_warps;
@@ -1081,7 +1085,7 @@ __global__ void Marlin(
 
     // Determine "position" inside the thread-block (based on warp and
     // thread-id)
-    int warp_id = threadIdx.x / 32;
+    auto warp_id = threadIdx.x / 32;
     int n_warps =
         thread_n_blocks / 4;  // Each warp processes 4 16-size tiles over N
 
@@ -1090,7 +1094,7 @@ __global__ void Marlin(
 
     cur_k += warp_row * 16;
 
-    int th_id = threadIdx.x % 32;
+    auto th_id = threadIdx.x % 32;
     cur_k += (th_id % 4) * 2;  // Due to tensor-core layout for fp16 B matrix
 
     int s_col_shift =
@@ -1155,7 +1159,7 @@ __global__ void Marlin(
               (reinterpret_cast<int*>(sh_zp_stage))[zp_sh_rd + i];
         }
       } else {
-        int warp_id = threadIdx.x / 32;
+        auto warp_id = threadIdx.x / 32;
         int n_warps = thread_n_blocks / 4;
 
         int warp_row = warp_id / n_warps;
@@ -1193,7 +1197,7 @@ __global__ void Marlin(
                                      (pipe / (group_blocks / thread_k_blocks)));
           reinterpret_cast<int4*>(&frag_zpf[k % 2])[0] = sh_zp_stage[zp_sh_rd];
         } else {
-          int warp_id = threadIdx.x / 32;
+          auto warp_id = threadIdx.x / 32;
           int n_warps = thread_n_blocks / 4;
 
           int warp_row = warp_id / n_warps;
@@ -1319,7 +1323,7 @@ __global__ void Marlin(
   auto thread_block_reduce = [&]() {
     constexpr int red_off = threads / b_sh_stride_threads / 2;
     if (red_off >= 1) {
-      int red_idx = threadIdx.x / b_sh_stride_threads;
+      auto red_idx = threadIdx.x / b_sh_stride_threads;
       constexpr int red_sh_stride = b_sh_stride_threads * 4 * 2;
       constexpr int red_sh_delta = b_sh_stride_threads;
       int red_sh_rd = red_sh_stride * (threadIdx.x / b_sh_stride_threads) +
@@ -1386,7 +1390,7 @@ __global__ void Marlin(
                     4 * (threadIdx.x / 32) + threadIdx.x % 4;
       c_gl_wr += (2 * thread_n_blocks) * slice_col;
       constexpr int c_sh_wr_delta = active_threads;
-      int c_sh_wr = threadIdx.x;
+      auto c_sh_wr = threadIdx.x;
 
       int row = (threadIdx.x % 32) / 4;
 
@@ -1780,8 +1784,8 @@ __global__ void Marlin(
                HAS_ZP, GROUP_BLOCKS, IS_ZP_FLOAT>                              \
             <<<blocks, NUM_THREADS, max_shared_mem, stream>>>(                 \
                 A_ptr, B_ptr, C_ptr, C_tmp_ptr, s_ptr, zp_ptr, g_idx_ptr,      \
-                num_groups, prob_m, prob_n, prob_k, locks, use_atomic_add,     \
-                use_fp32_reduce);                                              \
+                num_groups, prob_m, prob_n, prob_k, lda, locks,                \
+                use_atomic_add, use_fp32_reduce);                              \
       }                                                                        \
     }
 
@@ -2071,7 +2075,7 @@ exec_config_t determine_thread_config(int prob_m, int prob_n, int prob_k,
 template <typename scalar_t>
 void marlin_mm(const void* A, const void* B, void* C, void* C_tmp, void* s,
                void* zp, void* g_idx, void* perm, void* a_tmp, int prob_m,
-               int prob_n, int prob_k, void* workspace,
+               int prob_n, int prob_k, int lda, void* workspace,
                vllm::ScalarType const& q_type, bool has_act_order,
                bool is_k_full, bool has_zp, int num_groups, int group_size,
                int dev, cudaStream_t stream, int thread_k, int thread_n,
@@ -2184,8 +2188,9 @@ void marlin_mm(const void* A, const void* B, void* C, void* C_tmp, void* s,
     // Permute A columns
     int block_rows = div_ceil(prob_m, blocks);
     permute_cols_kernel<<<blocks, default_threads, 0, stream>>>(
-        A_ptr, perm_ptr, a_tmp_ptr, prob_m, prob_k, block_rows);
+        A_ptr, perm_ptr, a_tmp_ptr, prob_m, prob_k, lda, block_rows);
     A_ptr = a_tmp_ptr;
+    lda = prob_k;
   }
 
   // If we have a full K, then we can run the non-act-order version of Marlin
@@ -2244,7 +2249,7 @@ void marlin_mm(const void* A, const void* B, void* C, void* C_tmp, void* s,
                   ", num_bits = ", num_bits);
     }
 
-    A_ptr += 16 * thread_m_blocks * (prob_k / 8) * par;
+    A_ptr += 16 * thread_m_blocks * (lda / 8) * par;
     C_ptr += 16 * thread_m_blocks * (prob_n / 8) * par;
   }
 }
@@ -2300,7 +2305,10 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
 
   // Verify device and strides
   TORCH_CHECK(a.device().is_cuda(), "A is not on GPU");
-  TORCH_CHECK(a.is_contiguous(), "A is not contiguous");
+  TORCH_CHECK(a.stride(1) == 1, "A.stride(1) is not 1");
+  // We use int4 (16 bytes) to load A, so A must aligned to 16 bytes
+  TORCH_CHECK(a.stride(0) % 8 == 0, "A.stride(0) must divisible by 8");
+  TORCH_CHECK(((uint64_t)a.data_ptr()) % 16 == 0, "A must aligned to 16 bytes");
 
   TORCH_CHECK(b_q_weight.device().is_cuda(), "b_q_weight is not on GPU");
   TORCH_CHECK(b_q_weight.is_contiguous(), "b_q_weight is not contiguous");
@@ -2432,7 +2440,7 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
         a.data_ptr<at::Half>(), b_q_weight.data_ptr(), c.data_ptr<at::Half>(),
         c_tmp.data_ptr<float>(), b_scales.data_ptr<at::Half>(),
         b_zeros.data_ptr(), g_idx.data_ptr(), perm.data_ptr(),
-        a_tmp.data_ptr<at::Half>(), size_m, size_n, size_k,
+        a_tmp.data_ptr<at::Half>(), size_m, size_n, size_k, a.stride(0),
         workspace.data_ptr(), b_q_type, has_act_order, is_k_full, has_zp,
         num_groups, group_size, dev, at::cuda::getCurrentCUDAStream(dev),
         thread_k, thread_n, sms, marlin::max_par, use_atomic_add,
@@ -2443,10 +2451,10 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
         c.data_ptr<at::BFloat16>(), c_tmp.data_ptr<float>(),
         b_scales.data_ptr<at::BFloat16>(), b_zeros.data_ptr(), g_idx.data_ptr(),
         perm.data_ptr(), a_tmp.data_ptr<at::BFloat16>(), size_m, size_n, size_k,
-        workspace.data_ptr(), b_q_type, has_act_order, is_k_full, has_zp,
-        num_groups, group_size, dev, at::cuda::getCurrentCUDAStream(dev),
-        thread_k, thread_n, sms, marlin::max_par, use_atomic_add,
-        use_fp32_reduce, is_zp_float);
+        a.stride(0), workspace.data_ptr(), b_q_type, has_act_order, is_k_full,
+        has_zp, num_groups, group_size, dev,
+        at::cuda::getCurrentCUDAStream(dev), thread_k, thread_n, sms,
+        marlin::max_par, use_atomic_add, use_fp32_reduce, is_zp_float);
   } else {
     TORCH_CHECK(false, "gpt_marlin_gemm only supports bfloat16 and float16");
   }

csrc/quantization/gptq_marlin/gptq_marlin_repack.cu

@@ -15,7 +15,7 @@ __global__ void gptq_marlin_repack_kernel(
   int n_tiles = size_n / tile_n_size;
   int block_k_tiles = div_ceil(k_tiles, gridDim.x);
 
-  int start_k_tile = blockIdx.x * block_k_tiles;
+  auto start_k_tile = blockIdx.x * block_k_tiles;
   if (start_k_tile >= k_tiles) {
     return;
   }
@@ -71,8 +71,8 @@ __global__ void gptq_marlin_repack_kernel(
 
     if constexpr (has_perm) {
       if (threadIdx.x < stage_size) {
-        int k_id = threadIdx.x / stage_n_threads;
-        int n_id = threadIdx.x % stage_n_threads;
+        auto k_id = threadIdx.x / stage_n_threads;
+        auto n_id = threadIdx.x % stage_n_threads;
 
         uint32_t const* sh_perm_int_ptr =
             reinterpret_cast<uint32_t const*>(sh_perm_ptr);
@@ -88,8 +88,8 @@ __global__ void gptq_marlin_repack_kernel(
 
     } else {
       if (threadIdx.x < stage_size) {
-        int k_id = threadIdx.x / stage_n_threads;
-        int n_id = threadIdx.x % stage_n_threads;
+        auto k_id = threadIdx.x / stage_n_threads;
+        auto n_id = threadIdx.x % stage_n_threads;
 
         int first_k = k_tile_id * tile_k_size;
         int first_k_packed = first_k / pack_factor;
@@ -109,8 +109,8 @@ __global__ void gptq_marlin_repack_kernel(
       return;
     }
 
-    int warp_id = threadIdx.x / 32;
-    int th_id = threadIdx.x % 32;
+    auto warp_id = threadIdx.x / 32;
+    auto th_id = threadIdx.x % 32;
 
     if (warp_id >= 4) {
       return;
@@ -339,4 +339,4 @@ TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, CUDA, m) {
 
 TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, Meta, m) {
   m.impl("gptq_marlin_repack", &gptq_marlin_repack_meta);
-}
+}

csrc/quantization/marlin/dense/marlin_cuda_kernel.cu

@@ -277,12 +277,12 @@ __global__ void Marlin(
       b_gl_stride * (threadIdx.x / b_sh_stride) + (threadIdx.x % b_sh_stride);
   b_gl_rd += b_sh_stride * slice_col;
   b_gl_rd += b_gl_rd_delta_o * slice_row;
-  int b_sh_wr = threadIdx.x;
-  int b_sh_rd = threadIdx.x;
+  auto b_sh_wr = threadIdx.x;
+  auto b_sh_rd = threadIdx.x;
 
   int s_gl_rd = s_gl_stride * ((thread_k_blocks * slice_row) / group_blocks) +
                 s_sh_stride * slice_col + threadIdx.x;
-  int s_sh_wr = threadIdx.x;
+  auto s_sh_wr = threadIdx.x;
   int s_sh_rd;
   // We use a different scale layout for grouped and column-wise quantization as
   // we scale a `half2` tile in column-major layout in the former and in
@@ -455,7 +455,7 @@ __global__ void Marlin(
   auto thread_block_reduce = [&]() {
     constexpr int red_off = threads / b_sh_stride / 2;
     if (red_off >= 1) {
-      int red_idx = threadIdx.x / b_sh_stride;
+      auto red_idx = threadIdx.x / b_sh_stride;
       constexpr int red_sh_stride = b_sh_stride * 4 * 2;
       constexpr int red_sh_delta = b_sh_stride;
       int red_sh_rd = red_sh_stride * (threadIdx.x / b_sh_stride) +
@@ -522,7 +522,7 @@ __global__ void Marlin(
                     4 * (threadIdx.x / 32) + threadIdx.x % 4;
       c_gl_wr += (2 * thread_n_blocks) * slice_col;
       constexpr int c_sh_wr_delta = active_threads;
-      int c_sh_wr = threadIdx.x;
+      auto c_sh_wr = threadIdx.x;
 
       int row = (threadIdx.x % 32) / 4;
 

csrc/quantization/marlin/qqq/marlin_qqq_gemm_kernel.cu

@@ -353,10 +353,10 @@ __global__ void Marlin(
       b_gl_stride * (threadIdx.x / b_sh_stride) + (threadIdx.x % b_sh_stride);
   b_gl_rd += b_sh_stride * slice_col;
   b_gl_rd += b_gl_rd_delta_o * slice_row;
-  int b_sh_wr = threadIdx.x;
-  int b_sh_rd = threadIdx.x;
+  auto b_sh_wr = threadIdx.x;
+  auto b_sh_rd = threadIdx.x;
 
-  int s_tok_gl_rd = threadIdx.x;
+  auto s_tok_gl_rd = threadIdx.x;
   // NOTE(HandH1998): activation scale s_tok need shuffle to [0, 8, 1, 9, 2, 10,
   // 3, 11, 4, 12, 5, 13, 6, 14, 7, 15] for example, 0, 8 row scales serve for
   // thread 0, 1, 2, 3. For more details, refer to mma operand A layout as
@@ -368,8 +368,8 @@ __global__ void Marlin(
   int s_tok_sh_rd = (threadIdx.x % 32) / 4;
   bool s_tok_sh_wr_pred = threadIdx.x < prob_m;
 
-  int s_ch_gl_rd = s_ch_sh_stride * slice_col + threadIdx.x;
-  int s_ch_sh_wr = threadIdx.x;
+  auto s_ch_gl_rd = s_ch_sh_stride * slice_col + threadIdx.x;
+  auto s_ch_sh_wr = threadIdx.x;
   int s_ch_sh_rd = 16 * ((threadIdx.x / 32) % (thread_n_blocks / 4)) +
                    2 * ((threadIdx.x % 32) % 4);
   bool s_ch_sh_wr_pred = threadIdx.x < s_ch_sh_stride;
@@ -558,7 +558,7 @@ __global__ void Marlin(
   auto thread_block_reduce = [&]() {
     constexpr int red_off = threads / b_sh_stride / 2;
     if (red_off >= 1) {
-      int red_idx = threadIdx.x / b_sh_stride;
+      auto red_idx = threadIdx.x / b_sh_stride;
       constexpr int red_sh_stride = b_sh_stride * 4 * 2;
       constexpr int red_sh_delta = b_sh_stride;
       int red_sh_rd = red_sh_stride * (threadIdx.x / b_sh_stride) +
@@ -628,7 +628,7 @@ __global__ void Marlin(
                     8 * (threadIdx.x / 32) + (threadIdx.x % 4) * 2;
       c_gl_wr += (4 * thread_n_blocks) * slice_col;
       constexpr int c_sh_wr_delta = active_threads * 2;
-      int c_sh_wr = 2 * threadIdx.x;
+      auto c_sh_wr = 2 * threadIdx.x;
 
       int row = (threadIdx.x % 32) / 4;
 

csrc/quantization/marlin/sparse/marlin_24_cuda_kernel.cu

@@ -273,15 +273,15 @@ __global__ void Marlin_24(
                 (threadIdx.x % b_sh_stride_threads) * b_thread_vecs;
   b_gl_rd += b_sh_stride * slice_col;
   b_gl_rd += b_gl_rd_delta_o * slice_row;
-  int b_sh_wr = threadIdx.x * b_thread_vecs;
-  int b_sh_rd = threadIdx.x * b_thread_vecs;
+  auto b_sh_wr = threadIdx.x * b_thread_vecs;
+  auto b_sh_rd = threadIdx.x * b_thread_vecs;
 
   int m_gl_rd = m_gl_stride * (threadIdx.x / (m_sh_stride)) +
                 (threadIdx.x % (m_sh_stride));
   m_gl_rd += (m_sh_stride)*slice_col;
   m_gl_rd += m_gl_rd_delta_o * slice_row;
-  int m_sh_wr = threadIdx.x;
-  int m_sh_rd = threadIdx.x % 16 + (threadIdx.x / 32) * 16;
+  auto m_sh_wr = threadIdx.x;
+  auto m_sh_rd = threadIdx.x % 16 + (threadIdx.x / 32) * 16;
 
   int s_gl_rd;
   if constexpr (group_blocks == -1) {
@@ -291,7 +291,7 @@ __global__ void Marlin_24(
               s_sh_stride * slice_col + threadIdx.x;
   }
 
-  int s_sh_wr = threadIdx.x;
+  auto s_sh_wr = threadIdx.x;
   int s_sh_rd;
   // We use a different scale layout for grouped and column-wise quantization as
   // we scale a `half2` tile in column-major layout in the former and in
@@ -516,7 +516,7 @@ __global__ void Marlin_24(
   auto thread_block_reduce = [&]() {
     constexpr int red_off = threads / b_sh_stride_threads / 2;
     if (red_off >= 1) {
-      int red_idx = threadIdx.x / b_sh_stride_threads;
+      auto red_idx = threadIdx.x / b_sh_stride_threads;
       constexpr int red_sh_stride = b_sh_stride_threads * 4 * 2;
       constexpr int red_sh_delta = b_sh_stride_threads;
       int red_sh_rd = red_sh_stride * (threadIdx.x / b_sh_stride_threads) +
@@ -583,7 +583,7 @@ __global__ void Marlin_24(
                     8 * (threadIdx.x / 32) + (threadIdx.x % 32) / 4;
       c_gl_wr += (2 * thread_n_blocks) * slice_col;
       constexpr int c_sh_wr_delta = active_threads;
-      int c_sh_wr = threadIdx.x;
+      auto c_sh_wr = threadIdx.x;
 
       int col = 2 * ((threadIdx.x % 32) % 4);
 

csrc/rocm/attention.cu

@@ -284,18 +284,18 @@ __launch_bounds__(NUM_THREADS, 5) void paged_attention_ll4mi_QKV_mfma16_kernel(
     int max_ctx_blocks, const float* k_scale, const float* v_scale) {
   // clang-format on
   constexpr int NWARPS = NUM_THREADS / WARP_SIZE;
-  const int warpid = threadIdx.x / WARP_SIZE;
-  const int laneid = threadIdx.x % WARP_SIZE;
+  const auto warpid = threadIdx.x / WARP_SIZE;
+  const auto laneid = threadIdx.x % WARP_SIZE;
   const int lane4id = laneid % 4;
   const int lane16id = laneid % 16;
   const int rowid = laneid / 16;
 
-  const int seq_idx = blockIdx.x;
-  const int partition_idx = blockIdx.y;
+  const auto seq_idx = blockIdx.x;
+  const auto partition_idx = blockIdx.y;
 
   constexpr int T_PAR_SIZE = 256;  // token partition size set to 256
 
-  const int max_num_partitions = gridDim.y;
+  const auto max_num_partitions = gridDim.y;
 
   const int context_len = context_lens[seq_idx];
 
@@ -346,9 +346,9 @@ __launch_bounds__(NUM_THREADS, 5) void paged_attention_ll4mi_QKV_mfma16_kernel(
   // can be interpreted as B8x16 for 8 bit types
   _B16x8 Klocal[TLOOP][QKHELOOP];
 
-  const int wg_start_head_idx = blockIdx.z * GQA_RATIO;
-  const int wg_start_kv_head_idx = blockIdx.z;
-  const int total_num_heads = gridDim.z * GQA_RATIO;
+  const auto wg_start_head_idx = blockIdx.z * GQA_RATIO;
+  const auto wg_start_kv_head_idx = blockIdx.z;
+  const auto total_num_heads = gridDim.z * GQA_RATIO;
 
   // for QK mfma, tokens in multiples of TOKENS_PER_WARP are spread across warps
   // each mfma takes QH16xT16x16HE across warp
@@ -789,14 +789,14 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_QKV_mfma4_kernel(
     int max_ctx_blocks, const float* k_scale, const float* v_scale) {
   // clang-format on
   constexpr int NWARPS = NUM_THREADS / WARP_SIZE;
-  const int warpid = threadIdx.x / WARP_SIZE;
-  const int laneid = threadIdx.x % WARP_SIZE;
+  const auto warpid = threadIdx.x / WARP_SIZE;
+  const auto laneid = threadIdx.x % WARP_SIZE;
   const int lane4id = laneid % 4;
 
-  const int seq_idx = blockIdx.x;
-  const int partition_idx = blockIdx.y;
-  const int partition_size = blockDim.x;
-  const int max_num_partitions = gridDim.y;
+  const auto seq_idx = blockIdx.x;
+  const auto partition_idx = blockIdx.y;
+  const auto partition_size = blockDim.x;
+  const auto max_num_partitions = gridDim.y;
 
   const int context_len = context_lens[seq_idx];
   const int partition_start_token_idx = partition_idx * partition_size;
@@ -838,8 +838,8 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_QKV_mfma4_kernel(
     qk_max[h] = -FLT_MAX;
   }
 
-  const int wg_start_head_idx = blockIdx.z * GQA_RATIO;
-  const int wg_start_kv_head_idx = blockIdx.z;
+  const auto wg_start_head_idx = blockIdx.z * GQA_RATIO;
+  const auto wg_start_kv_head_idx = blockIdx.z;
 
   const int warp_start_token_idx =
       partition_start_token_idx + warpid * WARP_SIZE;
@@ -857,7 +857,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_QKV_mfma4_kernel(
 
     const int* block_table = block_tables + seq_idx * max_num_blocks_per_seq;
     // token id within partition
-    const int local_token_idx = threadIdx.x;
+    const auto local_token_idx = threadIdx.x;
     // token id within sequence
     const int global_token_idx = partition_start_token_idx + local_token_idx;
 
@@ -1126,7 +1126,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_QKV_mfma4_kernel(
 
   __syncthreads();
 
-  const int num_heads = gridDim.z * GQA_RATIO;
+  const auto num_heads = gridDim.z * GQA_RATIO;
   float* max_logits_ptr =
       max_logits + seq_idx * num_heads * max_num_partitions + partition_idx;
   float* exp_sums_ptr =
@@ -1268,14 +1268,14 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_reduce_kernel(
                                            // max_num_partitions, head_size]
     const int* __restrict__ context_lens,  // [num_seqs]
     const int max_num_partitions) {
-  const int num_heads = gridDim.x;
-  const int head_idx = blockIdx.x;
-  const int seq_idx = blockIdx.y;
+  const auto num_heads = gridDim.x;
+  const auto head_idx = blockIdx.x;
+  const auto seq_idx = blockIdx.y;
   const int context_len = context_lens[seq_idx];
   const int num_partitions = DIVIDE_ROUND_UP(context_len, PARTITION_SIZE);
   [[maybe_unused]] constexpr int NUM_WARPS = NUM_THREADS / WARP_SIZE;
-  const int warpid = threadIdx.x / WARP_SIZE;
-  [[maybe_unused]] const int laneid = threadIdx.x % WARP_SIZE;
+  const auto warpid = threadIdx.x / WARP_SIZE;
+  [[maybe_unused]] const auto laneid = threadIdx.x % WARP_SIZE;
 
   __shared__ float shared_global_exp_sum;
   // max num partitions supported is warp_size * NPAR_LOOPS
@@ -1294,7 +1294,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_reduce_kernel(
 
   #pragma unroll
     for (int i = 0; i < NPAR_LOOPS; i++) {
-      const int partition_no = i * WARP_SIZE + threadIdx.x;
+      const auto partition_no = i * WARP_SIZE + threadIdx.x;
       valid_partition[i] =
           (partition_no < num_partitions) ? partition_no : last_valid_partition;
     }
@@ -1324,7 +1324,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_reduce_kernel(
     }
   #pragma unroll
     for (int i = 0; i < NPAR_LOOPS; i++) {
-      const int partition_no = i * WARP_SIZE + threadIdx.x;
+      const auto partition_no = i * WARP_SIZE + threadIdx.x;
       rescaled_exp_sum[i] *= (partition_no < num_partitions)
                                  ? expf(reg_max_logit[i] - max_logit)
                                  : 0.0f;
@@ -1336,7 +1336,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_reduce_kernel(
     }
   #pragma unroll
     for (int i = 0; i < NPAR_LOOPS; i++) {
-      const int partition_no = i * WARP_SIZE + threadIdx.x;
+      const auto partition_no = i * WARP_SIZE + threadIdx.x;
       shared_exp_sums[partition_no] = rescaled_exp_sum[i];
     }
 

csrc/torch_bindings.cpp

@@ -365,6 +365,35 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   ops.def("cutlass_scaled_mm_supports_fp8(int cuda_device_capability) -> bool");
   ops.impl("cutlass_scaled_mm_supports_fp8", &cutlass_scaled_mm_supports_fp8);
 
+  // Check if cutlass grouped gemm is supported for CUDA devices of the given
+  // capability
+  ops.def("cutlass_group_gemm_supported(int cuda_device_capability) -> bool");
+  ops.impl("cutlass_group_gemm_supported", &cutlass_group_gemm_supported);
+
+  // CUTLASS w8a8 grouped GEMM
+  ops.def(
+      "cutlass_moe_mm(Tensor! out_tensors, Tensor a_tensors, Tensor b_tensors, "
+      "               Tensor a_scales, Tensor b_scales, Tensor expert_offsets, "
+      "               Tensor problem_sizes, Tensor a_strides, "
+      "               Tensor b_strides, Tensor c_strides) -> ()",
+      {stride_tag});
+  ops.impl("cutlass_moe_mm", torch::kCUDA, &cutlass_moe_mm);
+
+  // A function that computes data required to run fused MoE with w8a8 grouped
+  // GEMM. It takes topk_ids as an input, and computes expert_offsets
+  // (token start indices of each expert). In addition to this, it computes
+  // problem sizes for each expert's multiplication used by the two mms called
+  // from fused MoE operation, and arrays with permutations required to shuffle
+  // and de-shuffle the input/output of the fused operation.
+  ops.def(
+      "get_cutlass_moe_mm_data(Tensor topk_ids, Tensor! expert_offsets, "
+      "                        Tensor! problem_sizes1, Tensor! problem_sizes2, "
+      "                        Tensor! input_permutation, "
+      "                        Tensor! output_permutation, int num_experts, "
+      "                        int n, int k) -> ()",
+      {stride_tag});
+  ops.impl("get_cutlass_moe_mm_data", torch::kCUDA, &get_cutlass_moe_mm_data);
+
   // Check if cutlass scaled_mm supports block quantization (used by DeepSeekV3)
   ops.def(
       "cutlass_scaled_mm_supports_block_fp8(int cuda_device_capability) -> "

docs/source/_static/custom.js

@@ -10,8 +10,8 @@ document.addEventListener("DOMContentLoaded", function () {
     script.setAttribute("runllm-keyboard-shortcut", "Mod+j"); // cmd-j or ctrl-j to open the widget.
     script.setAttribute("runllm-name", "vLLM");
     script.setAttribute("runllm-position", "BOTTOM_RIGHT");
-    script.setAttribute("runllm-position-y", "20%");
-    script.setAttribute("runllm-position-x", "3%");
+    script.setAttribute("runllm-position-y", "120px");
+    script.setAttribute("runllm-position-x", "20px");
     script.setAttribute("runllm-assistant-id", "207");
   
     script.async = true;

docs/source/conf.py

@@ -85,6 +85,7 @@ html_static_path = ["_static"]
 html_js_files = ["custom.js"]
 html_css_files = ["custom.css"]
 
+myst_heading_anchors = 2
 myst_url_schemes = {
     'http': None,
     'https': None,
@@ -102,6 +103,11 @@ myst_url_schemes = {
         "title": "Pull Request #{{path}}",
         "classes": ["github"],
     },
+    "gh-project": {
+        "url": "https://github.com/vllm-project/projects/{{path}}",
+        "title": "Project #{{path}}",
+        "classes": ["github"],
+    },
     "gh-dir": {
         "url": "https://github.com/vllm-project/vllm/tree/main/{{path}}",
         "title": "{{path}}",

docs/source/contributing/overview.md

@@ -11,6 +11,15 @@ We also believe in the power of community support; thus, answering queries, offe
 
 Finally, one of the most impactful ways to support us is by raising awareness about vLLM. Talk about it in your blog posts and highlight how it's driving your incredible projects. Express your support on social media if you're using vLLM, or simply offer your appreciation by starring our repository!
 
+## Job Board
+
+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)
+- [New model requests](https://github.com/vllm-project/vllm/issues?q=is%3Aissue%20state%3Aopen%20label%3A%22new%20model%22)
+  - [Models with multi-modal capabilities](gh-project:10)
+
 ## License
 
 See <gh-file:LICENSE>.

docs/source/deployment/docker.md

@@ -34,11 +34,11 @@ If you need to use those dependencies (having accepted the license terms),
 create a custom Dockerfile on top of the base image with an extra layer that installs them:
 
 ```Dockerfile
-FROM vllm/vllm-openai:v0.8.0
+FROM vllm/vllm-openai:v0.8.2
 
 # e.g. install the `audio` and `video` optional dependencies
 # NOTE: Make sure the version of vLLM matches the base image!
-RUN uv pip install vllm[audio,video]==0.8.0
+RUN uv pip install --system vllm[audio,video]==0.8.2
 ```
 
 :::
@@ -52,7 +52,7 @@ with an extra layer that installs their code from source:
 ```Dockerfile
 FROM vllm/vllm-openai:latest
 
-RUN uv pip install git+https://github.com/huggingface/transformers.git
+RUN uv pip install --system git+https://github.com/huggingface/transformers.git
 ```
 
 :::

docs/source/deployment/k8s.md

@@ -4,6 +4,9 @@
 
 Deploying vLLM on Kubernetes is a scalable and efficient way to serve machine learning models. This guide walks you through deploying vLLM using native Kubernetes.
 
+* [Deployment with CPUs](#deployment-with-cpus)
+* [Deployment with GPUs](#deployment-with-gpus)
+
 Alternatively, you can deploy vLLM to Kubernetes using any of the following:
 * [Helm](frameworks/helm.md)
 * [InftyAI/llmaz](integrations/llmaz.md)
@@ -14,11 +17,107 @@ Alternatively, you can deploy vLLM to Kubernetes using any of the following:
 * [vllm-project/aibrix](https://github.com/vllm-project/aibrix)
 * [vllm-project/production-stack](integrations/production-stack.md)
 
-## Pre-requisite
+## Deployment with CPUs
 
-Ensure that you have a running [Kubernetes cluster with GPUs](https://kubernetes.io/docs/tasks/manage-gpus/scheduling-gpus/).
+:::{note}
+The use of CPUs here is for demonstration and testing purposes only and its performance will not be on par with GPUs.
+:::
 
-## Deployment using native K8s
+First, create a Kubernetes PVC and Secret for downloading and storing Hugging Face model:
+
+```bash
+cat <<EOF |kubectl apply -f -
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: vllm-models
+spec:
+  accessModes:
+    - ReadWriteOnce
+  volumeMode: Filesystem
+  resources:
+    requests:
+      storage: 50Gi
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: hf-token-secret
+type: Opaque
+data:
+  token: $(HF_TOKEN)
+```
+
+Next, start the vLLM server as a Kubernetes Deployment and Service:
+
+```bash
+cat <<EOF |kubectl apply -f -
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: vllm-server
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: vllm
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: vllm
+    spec:
+      containers:
+      - name: vllm
+        image: vllm/vllm-openai:latest
+        command: ["/bin/sh", "-c"]
+        args: [
+          "vllm serve meta-llama/Llama-3.2-1B-Instruct"
+        ]
+        env:
+        - name: HUGGING_FACE_HUB_TOKEN
+          valueFrom:
+            secretKeyRef:
+              name: hf-token-secret
+              key: token
+        ports:
+          - containerPort: 8000
+        volumeMounts:
+          - name: llama-storage
+            mountPath: /root/.cache/huggingface
+      volumes:
+      - name: llama-storage
+        persistentVolumeClaim:
+          claimName: vllm-models
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: vllm-server
+spec:
+  selector:
+    app.kubernetes.io/name: vllm
+  ports:
+  - protocol: TCP
+    port: 8000
+    targetPort: 8000
+  type: ClusterIP
+EOF
+```
+
+We can verify that the vLLM server has started successfully via the logs (this might take a couple of minutes to download the model):
+
+```console
+kubectl logs -l app.kubernetes.io/name=vllm
+...
+INFO:     Started server process [1]
+INFO:     Waiting for application startup.
+INFO:     Application startup complete.
+INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
+```
+
+## Deployment with GPUs
+
+**Pre-requisite**: Ensure that you have a running [Kubernetes cluster with GPUs](https://kubernetes.io/docs/tasks/manage-gpus/scheduling-gpus/).
 
 1. Create a PVC, Secret and Deployment for vLLM
 

docs/source/design/v1/prefix_caching.md

@@ -15,12 +15,13 @@ Block 3: |<------------------ prefix -------------------->| |<--- block tokens -
 In the example above, the KV cache in the first block can be uniquely identified with the token “A gentle breeze stirred”. The third block can be uniquely identified with the tokens in the block “laughed in the distance”, along with the prefix tokens “A gentle breeze stirred the leaves as children”. Therefore, we can build the block hash of `hash(tuple[components])`, where components are:
 
 * Parent hash value: The hash value of the parent hash block.
-* Block tokens: A tuple of tokens in this block. The reason to include the exact tokens is to reduce potential hash value collision.  
+* Block tokens: A tuple of tokens in this block. The reason to include the exact tokens is to reduce potential hash value collision.
 * Extra hashes: Other values required to make this block unique, such as LoRA IDs and multi-modality input hashes (see the example below).
 
-Note 1: We only cache full blocks.
+> **Note 1:** We only cache full blocks.
 
-Note 2: The above hash key structure is not 100% collision free. Theoretically it’s still possible for the different prefix tokens to have the same hash value, but this should be nearly impossible to happen. Of course, contributions are welcome if you have an awesome idea to eliminate collusion entirely.
+> **Note 2:** The above hash key structure is not 100% collision free. Theoretically it’s still possible for the different prefix tokens to have the same hash value. To avoid any hash collisions **in a multi-tenant setup, we advise to use SHA256** as hash function instead of the default builtin hash.
+SHA256 is supported since vLLM v0.8.3 and must be enabled with a command line argument. It comes with a performance impact of about 100-200ns per token (~6ms for 50k tokens of context).
 
 **A hashing example with multi-modality inputs**  
 In this example, we illustrate how prefix caching works with multi-modality inputs (e.g., images). Assuming we have a request with the following messages:

docs/source/features/quantization/bnb.md

@@ -9,7 +9,7 @@ Compared to other quantization methods, BitsAndBytes eliminates the need for cal
 Below are the steps to utilize BitsAndBytes with vLLM.
 
 ```console
-pip install bitsandbytes>=0.45.0
+pip install bitsandbytes>=0.45.3
 ```
 
 vLLM reads the model's config file and supports both in-flight quantization and pre-quantized checkpoint.

docs/source/features/reasoning_outputs.md

@@ -4,16 +4,19 @@
 
 vLLM offers support for reasoning models like [DeepSeek R1](https://huggingface.co/deepseek-ai/DeepSeek-R1), which are designed to generate outputs containing both reasoning steps and final conclusions.
 
-Reasoning models return a additional `reasoning_content` field in their outputs, which contains the reasoning steps that led to the final conclusion. This field is not present in the outputs of other models.
+Reasoning models return an additional `reasoning_content` field in their outputs, which contains the reasoning steps that led to the final conclusion. This field is not present in the outputs of other models.
 
 ## Supported Models
 
 vLLM currently supports the following reasoning models:
 
-| Model Series | Parser Name | Structured Output Support |
-|--------------|-------------|------------------|
-| [DeepSeek R1 series](https://huggingface.co/collections/deepseek-ai/deepseek-r1-678e1e131c0169c0bc89728d) | `deepseek_r1` | `guided_json`, `guided_regex` |
-| [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) | `deepseek_r1` | `guided_json`, `guided_regex` |
+| Model Series | Parser Name | Structured Output Support | Tool Calling |
+|--------------|-------------|------------------|-------------|
+| [DeepSeek R1 series](https://huggingface.co/collections/deepseek-ai/deepseek-r1-678e1e131c0169c0bc89728d) | `deepseek_r1` | `guided_json`, `guided_regex` | ❌ |
+| [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) | `deepseek_r1` | `guided_json`, `guided_regex` | ✅ |
+| [IBM Granite 3.2 language models](https://huggingface.co/collections/ibm-granite/granite-32-language-models-67b3bc8c13508f6d064cff9a) | `granite` | ❌ | ❌ |
+
+- IBM Granite 3.2 reasoning is disabled by default; to enable it, you must also pass `thinking=True` in your `chat_template_kwargs`.
 
 ## Quickstart
 
@@ -43,6 +46,7 @@ model = models.data[0].id
 
 # Round 1
 messages = [{"role": "user", "content": "9.11 and 9.8, which is greater?"}]
+# For granite, add: `extra_body={"chat_template_kwargs": {"thinking": True}}`
 response = client.chat.completions.create(model=model, messages=messages)
 
 reasoning_content = response.choices[0].message.reasoning_content
@@ -97,6 +101,7 @@ models = client.models.list()
 model = models.data[0].id
 
 messages = [{"role": "user", "content": "9.11 and 9.8, which is greater?"}]
+# For granite, add: `extra_body={"chat_template_kwargs": {"thinking": True}}`
 stream = client.chat.completions.create(model=model,
                                         messages=messages,
                                         stream=True)
@@ -170,10 +175,51 @@ print("reasoning_content: ", completion.choices[0].message.reasoning_content)
 print("content: ", completion.choices[0].message.content)
 ```
 
+## Tool Calling
+
+The reasoning content is also available when both tool calling and the reasoning parser are enabled. Additionally, tool calling only parses functions from the `content` field, not from the `reasoning_content`.
+
+```python
+from openai import OpenAI
+
+client = OpenAI(base_url="http://localhost:8000/v1", api_key="dummy")
+
+tools = [{
+    "type": "function",
+    "function": {
+        "name": "get_weather",
+        "description": "Get the current weather in a given location",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "location": {"type": "string", "description": "City and state, e.g., 'San Francisco, CA'"},
+                "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}
+            },
+            "required": ["location", "unit"]
+        }
+    }
+}]
+
+response = client.chat.completions.create(
+    model=client.models.list().data[0].id,
+    messages=[{"role": "user", "content": "What's the weather like in San Francisco?"}],
+    tools=tools,
+    tool_choice="auto"
+)
+
+print(response)
+tool_call = response.choices[0].message.tool_calls[0].function
+
+print(f"reasoning_content: {response.choices[0].message.reasoning_content}")
+print(f"Function called: {tool_call.name}")
+print(f"Arguments: {tool_call.arguments}")
+```
+
+For more examples, please refer to <gh-file:examples/online_serving/openai_chat_completion_tool_calls_with_reasoning.py> .
+
 ## Limitations
 
 - The reasoning content is only available for online serving's chat completion endpoint (`/v1/chat/completions`).
-- It is not compatible with [`tool_calling`](#tool_calling).
 
 ## How to support a new reasoning model
 

docs/source/features/spec_decode.md

@@ -30,8 +30,10 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
 llm = LLM(
     model="facebook/opt-6.7b",
     tensor_parallel_size=1,
-    speculative_model="facebook/opt-125m",
-    num_speculative_tokens=5,
+    speculative_config={
+        "model": "facebook/opt-125m",
+        "num_speculative_tokens": 5,
+    },
 )
 outputs = llm.generate(prompts, sampling_params)
 
@@ -45,10 +47,14 @@ To perform the same with an online mode launch the server:
 
 ```bash
 python -m vllm.entrypoints.openai.api_server --host 0.0.0.0 --port 8000 --model facebook/opt-6.7b \
-    --seed 42 -tp 1 --speculative_model facebook/opt-125m \
-    --num_speculative_tokens 5 --gpu_memory_utilization 0.8
+    --seed 42 -tp 1 --gpu_memory_utilization 0.8 \
+    --speculative_config '{"model": "facebook/opt-125m", "num_speculative_tokens": 5}'
 ```
 
+:::{warning}
+Note: Please use `--speculative_config` to set all configurations related to speculative decoding. The previous method of specifying the model through `--speculative_model` and adding related parameters (e.g., `--num_speculative_tokens`) separately will be deprecated in the next release.
+:::
+
 Then use a client:
 
 ```python
@@ -101,9 +107,11 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
 llm = LLM(
     model="facebook/opt-6.7b",
     tensor_parallel_size=1,
-    speculative_model="[ngram]",
-    num_speculative_tokens=5,
-    ngram_prompt_lookup_max=4,
+    speculative_config={
+        "method": "ngram",
+        "num_speculative_tokens": 5,
+        "prompt_lookup_max": 4,
+    },
 )
 outputs = llm.generate(prompts, sampling_params)
 
@@ -131,8 +139,10 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
 llm = LLM(
     model="meta-llama/Meta-Llama-3.1-70B-Instruct",
     tensor_parallel_size=4,
-    speculative_model="ibm-ai-platform/llama3-70b-accelerator",
-    speculative_draft_tensor_parallel_size=1,
+    speculative_config={
+        "model": "ibm-ai-platform/llama3-70b-accelerator",
+        "draft_tensor_parallel_size": 1,
+    },
 )
 outputs = llm.generate(prompts, sampling_params)
 
@@ -175,8 +185,10 @@ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
 llm = LLM(
     model="meta-llama/Meta-Llama-3-8B-Instruct",
     tensor_parallel_size=4,
-    speculative_model="yuhuili/EAGLE-LLaMA3-Instruct-8B",
-    speculative_draft_tensor_parallel_size=1,
+    speculative_config={
+        "model": "yuhuili/EAGLE-LLaMA3-Instruct-8B",
+        "draft_tensor_parallel_size": 1,
+    },
 )
 
 outputs = llm.generate(prompts, sampling_params)
@@ -194,11 +206,10 @@ A few important things to consider when using the EAGLE based draft models:
    be able to be loaded and used directly by vLLM after [PR 12304](https://github.com/vllm-project/vllm/pull/12304).
    If you are using vllm version before [PR 12304](https://github.com/vllm-project/vllm/pull/12304), please use the
    [script](https://gist.github.com/abhigoyal1997/1e7a4109ccb7704fbc67f625e86b2d6d) to convert the speculative model,
-   and specify `speculative_model="path/to/modified/eagle/model"`. If weight-loading problems still occur when using
-   the latest version of vLLM, please leave a comment or raise an issue.
+   and specify `"model": "path/to/modified/eagle/model"` in `speculative_config`. If weight-loading problems still occur when using the latest version of vLLM, please leave a comment or raise an issue.
 
 2. The EAGLE based draft models need to be run without tensor parallelism
-   (i.e. speculative_draft_tensor_parallel_size is set to 1), although
+   (i.e. draft_tensor_parallel_size is set to 1 in `speculative_config`), although
    it is possible to run the main model using tensor parallelism (see example above).
 
 3. When using EAGLE-based speculators with vLLM, the observed speedup is lower than what is

docs/source/getting_started/installation/cpu.md

@@ -193,7 +193,7 @@ vLLM CPU backend supports the following vLLM features:
 
 ## Related runtime environment variables
 
-- `VLLM_CPU_KVCACHE_SPACE`: specify the KV Cache size (e.g, `VLLM_CPU_KVCACHE_SPACE=40` means 40 GB space for KV cache), larger setting will allow vLLM running more requests in parallel. This parameter should be set based on the hardware configuration and memory management pattern of users.
+- `VLLM_CPU_KVCACHE_SPACE`: specify the KV Cache size (e.g, `VLLM_CPU_KVCACHE_SPACE=40` means 40 GiB space for KV cache), larger setting will allow vLLM running more requests in parallel. This parameter should be set based on the hardware configuration and memory management pattern of users.
 - `VLLM_CPU_OMP_THREADS_BIND`: specify the CPU cores dedicated to the OpenMP threads. For example, `VLLM_CPU_OMP_THREADS_BIND=0-31` means there will be 32 OpenMP threads bound on 0-31 CPU cores. `VLLM_CPU_OMP_THREADS_BIND=0-31|32-63` means there will be 2 tensor parallel processes, 32 OpenMP threads of rank0 are bound on 0-31 CPU cores, and the OpenMP threads of rank1 are bound on 32-63 CPU cores.
 - `VLLM_CPU_MOE_PREPACK`: whether to use prepack for MoE layer. This will be passed to `ipex.llm.modules.GatedMLPMOE`. Default is `1` (True). On unsupported CPUs, you might need to set this to `0` (False).
 

docs/source/getting_started/quickstart.md

@@ -58,6 +58,11 @@ from vllm import LLM, SamplingParams
 ```
 
 The next section defines a list of input prompts and sampling parameters for text generation. The [sampling temperature](https://arxiv.org/html/2402.05201v1) is set to `0.8` and the [nucleus sampling probability](https://en.wikipedia.org/wiki/Top-p_sampling) is set to `0.95`. You can find more information about the sampling parameters [here](#sampling-params).
+:::{important}
+By default, vLLM will use sampling parameters recommended by model creator by applying the `generation_config.json` from the Hugging Face model repository if it exists. In most cases, this will provide you with the best results by default if {class}`~vllm.SamplingParams` is not specified.
+
+However, if vLLM's default sampling parameters are preferred, please set `generation_config="vllm"` when creating the {class}`~vllm.LLM` instance.
+:::
 
 ```python
 prompts = [
@@ -76,7 +81,7 @@ llm = LLM(model="facebook/opt-125m")
 ```
 
 :::{note}
-By default, vLLM downloads models from [HuggingFace](https://huggingface.co/). If you would like to use models from [ModelScope](https://www.modelscope.cn), set the environment variable `VLLM_USE_MODELSCOPE` before initializing the engine.
+By default, vLLM downloads models from [Hugging Face](https://huggingface.co/). If you would like to use models from [ModelScope](https://www.modelscope.cn), set the environment variable `VLLM_USE_MODELSCOPE` before initializing the engine.
 :::
 
 Now, the fun part! The outputs are generated using `llm.generate`. It adds the input prompts to the vLLM engine's waiting queue and executes the vLLM engine to generate the outputs with high throughput. The outputs are returned as a list of `RequestOutput` objects, which include all of the output tokens.
@@ -107,6 +112,11 @@ vllm serve Qwen/Qwen2.5-1.5B-Instruct
 By default, the server uses a predefined chat template stored in the tokenizer.
 You can learn about overriding it [here](#chat-template).
 :::
+:::{important}
+By default, the server applies `generation_config.json` from the huggingface model repository if it exists. This means the default values of certain sampling parameters can be overridden by those recommended by the model creator.
+
+To disable this behavior, please pass `--generation-config vllm` when launching the server.
+:::
 
 This server can be queried in the same format as OpenAI API. For example, to list the models:
 

docs/source/getting_started/v1_user_guide.md

@@ -47,9 +47,9 @@ This living user guide outlines a few known **important changes and limitations*
 | **Logprobs Calculation**                    | <nobr>🟢 Functional</nobr>                                                        |
 | **LoRA**                                    | <nobr>🟢 Functional ([PR #13096](https://github.com/vllm-project/vllm/pull/13096))</nobr>|
 | **Multimodal Models**                       | <nobr>🟢 Functional</nobr>                                                        |
+| **FP8 KV Cache**                            | <nobr>🟢 Functional on Hopper devices ([PR #15191](https://github.com/vllm-project/vllm/pull/15191))</nobr>|
 | **Spec Decode**                             | <nobr>🚧 WIP ([PR #13933](https://github.com/vllm-project/vllm/pull/13933))</nobr>|
 | **Prompt Logprobs with Prefix Caching**     | <nobr>🟡 Planned ([RFC #13414](https://github.com/vllm-project/vllm/issues/13414))</nobr>|
-| **FP8 KV Cache**                            | <nobr>🟡 Planned</nobr>                                                           |
 | **Structured Output Alternative Backends**  | <nobr>🟡 Planned</nobr>                                                           |
 | **Embedding Models**                        | <nobr>🟡 Planned ([RFC #12249](https://github.com/vllm-project/vllm/issues/12249))</nobr> |
 | **Mamba Models**                            | <nobr>🟡 Planned</nobr>                                                           |
@@ -129,9 +129,10 @@ in progress.
 - **Spec Decode**: Currently, only ngram-based spec decode is supported in V1. There
   will be follow-up work to support other types of spec decode (e.g., see [PR #13933](https://github.com/vllm-project/vllm/pull/13933)). We will prioritize the support for Eagle, MTP compared to draft model based spec decode.
 
-#### Features to Be Supported
+- **Multimodal Models**: V1 is almost fully compatible with V0 except that interleaved modality input is not supported yet.
+  See [here](https://github.com/orgs/vllm-project/projects/8) for the status of upcoming features and optimizations.
 
-- **FP8 KV Cache**: While vLLM V1 introduces new FP8 kernels for model weight quantization, support for an FP8 key–value cache is not yet available. Users must continue using FP16 (or other supported precisions) for the KV cache.
+#### Features to Be Supported
 
 - **Structured Output Alternative Backends**: Structured output alternative backends (outlines, guidance) support is planned. V1 currently
   supports only the `xgrammar:no_fallback` mode, meaning that it will error out if the output schema is unsupported by xgrammar.
@@ -156,6 +157,9 @@ vLLM V1 is currently optimized for decoder-only transformers. Models requiring
 
 For a complete list of supported models, see the [list of supported models](https://docs.vllm.ai/en/latest/models/supported_models.html).
 
-## FAQ
+## Frequently Asked Questions
 
-TODO
+**I'm using vLLM V1 and I'm getting CUDA OOM errors. What should I do?**
+The default `max_num_seqs` has been raised from `256` in V0 to `1024` in V1. If you encounter CUDA OOM only when using V1 engine, try setting a lower value of `max_num_seqs` or `gpu_memory_utilization`.
+
+On the other hand, if you get an error about insufficient memory for the cache blocks, you should increase `gpu_memory_utilization` as this indicates that your GPU has sufficient memory but you're not allocating enough to vLLM for KV cache blocks.

docs/source/index.md

@@ -43,7 +43,7 @@ vLLM is flexible and easy to use with:
 - Tensor parallelism and pipeline parallelism support for distributed inference
 - Streaming outputs
 - OpenAI-compatible API server
-- Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs, Gaudi® accelerators and GPUs, PowerPC CPUs, TPU, and AWS Trainium and Inferentia Accelerators.
+- Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs, Gaudi® accelerators and GPUs, IBM Power CPUs, TPU, and AWS Trainium and Inferentia Accelerators.
 - Prefix caching support
 - Multi-lora support
 

docs/source/models/extensions/fastsafetensor.md

@@ -0,0 +1,5 @@
+Loading Model weights with fastsafetensors
+===================================================================
+
+Using fastsafetensor library enables loading model weights to GPU memory by leveraging GPU direct storage. See https://github.com/foundation-model-stack/fastsafetensors for more details.
+For enabling this feature, set the environment variable ``USE_FASTSAFETENSOR`` to ``true``

docs/source/models/extensions/index.md

@@ -5,4 +5,5 @@
 
 runai_model_streamer
 tensorizer
+fastsafetensor
 :::

docs/source/models/generative_models.md

@@ -46,6 +46,11 @@ for output in outputs:
     print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
 ```
 
+:::{important}
+By default, vLLM will use sampling parameters recommended by model creator by applying the `generation_config.json` from the huggingface model repository if it exists. In most cases, this will provide you with the best results by default if {class}`~vllm.SamplingParams` is not specified.
+
+However, if vLLM's default sampling parameters are preferred, please pass `generation_config="vllm"` when creating the {class}`~vllm.LLM` instance.
+:::
 A code example can be found here: <gh-file:examples/offline_inference/basic/basic.py>
 
 ### `LLM.beam_search`

docs/source/models/supported_models.md

@@ -57,10 +57,10 @@ llm = LLM(model=..., task="generate")  # Name or path of your model
 llm.apply_model(lambda model: print(type(model)))
 ```
 
-If it is `TransformersModel` then it means it's based on Transformers!
+If it is `TransformersForCausalLM` then it means it's based on Transformers!
 
 :::{tip}
-You can force the use of `TransformersModel` by setting `model_impl="transformers"` for <project:#offline-inference> or `--model-impl transformers` for the <project:#openai-compatible-server>.
+You can force the use of `TransformersForCausalLM` by setting `model_impl="transformers"` for <project:#offline-inference> or `--model-impl transformers` for the <project:#openai-compatible-server>.
 :::
 
 :::{note}
@@ -73,7 +73,7 @@ The Transformers fallback explicitly supports the following features:
 
 - <project:#quantization-index> (except GGUF)
 - <project:#lora-adapter>
-- <project:#distributed-serving> (pipeline parallel coming soon <gh-pr:12832>!)
+- <project:#distributed-serving> (requires `transformers>=4.49.0`)
 
 #### Remote code
 
@@ -119,7 +119,7 @@ Here is what happens in the background:
 
 1. The config is loaded
 2. `MyModel` Python class is loaded from the `auto_map`, and we check that the model `_supports_attention_backend`.
-3. The `TransformersModel` backend is used. See <gh-file:vllm/model_executor/models/transformers.py>, which leverage `self.config._attn_implementation = "vllm"`, thus the need to use `ALL_ATTENTION_FUNCTION`.
+3. The `TransformersForCausalLM` backend is used. See <gh-file:vllm/model_executor/models/transformers.py>, which leverage `self.config._attn_implementation = "vllm"`, thus the need to use `ALL_ATTENTION_FUNCTION`.
 
 To make your model compatible with tensor parallel, it needs:
 
@@ -836,14 +836,14 @@ See [this page](#generative-models) for more information on how to use generativ
   * `openbmb/MiniCPM-o-2_6`, etc.
   * ✅︎
   * ✅︎
-  *
+  * ✅︎
 - * `MiniCPMV`
   * MiniCPM-V
   * T + I<sup>E+</sup> + V<sup>E+</sup>
   * `openbmb/MiniCPM-V-2` (see note), `openbmb/MiniCPM-Llama3-V-2_5`, `openbmb/MiniCPM-V-2_6`, etc.
   * ✅︎
   * ✅︎
-  *
+  * ✅︎
 - * `MllamaForConditionalGeneration`
   * Llama 3.2
   * T + I<sup>+</sup>
@@ -853,7 +853,7 @@ See [this page](#generative-models) for more information on how to use generativ
   *
 - * `MolmoForCausalLM`
   * Molmo
-  * T + I
+  * T + I<sup>+</sup>
   * `allenai/Molmo-7B-D-0924`, `allenai/Molmo-7B-O-0924`, etc.
   * ✅︎
   * ✅︎

docs/source/serving/engine_args.md

@@ -2,7 +2,12 @@
 
 # Engine Arguments
 
-Below, you can find an explanation of every engine argument for vLLM:
+Engine arguments control the behavior of the vLLM engine.
+
+- For [offline inference](#offline-inference), they are part of the arguments to `LLM` class.
+- For [online serving](#openai-compatible-server), they are part of the arguments to `vllm serve`.
+
+Below, you can find an explanation of every engine argument:
 
 <!--- pyml disable-num-lines 7 no-space-in-emphasis -->
 ```{eval-rst}
@@ -15,7 +20,7 @@ Below, you can find an explanation of every engine argument for vLLM:
 
 ## Async Engine Arguments
 
-Below are the additional arguments related to the asynchronous engine:
+Additional arguments are available to the asynchronous engine which is used for online serving:
 
 <!--- pyml disable-num-lines 7 no-space-in-emphasis -->
 ```{eval-rst}

docs/source/serving/offline_inference.md

@@ -97,6 +97,13 @@ llm = LLM(model="adept/fuyu-8b",
           max_num_seqs=2)
 ```
 
+#### Adjust cache size
+
+If you run out of CPU RAM, try the following options:
+
+- (Multi-modal models only) you can set the size of multi-modal input cache using `VLLM_MM_INPUT_CACHE_GIB` environment variable (default 4 GiB).
+- (CPU backend only) you can set the size of KV cache using `VLLM_CPU_KVCACHE_SPACE` environment variable (default 4 GiB).
+
 ### Performance optimization and tuning
 
 You can potentially improve the performance of vLLM by finetuning various options.

docs/source/serving/openai_compatible_server.md

@@ -33,7 +33,11 @@ print(completion.choices[0].message)
 vLLM supports some parameters that are not supported by OpenAI, `top_k` for example.
 You can pass these parameters to vLLM using the OpenAI client in the `extra_body` parameter of your requests, i.e. `extra_body={"top_k": 50}` for `top_k`.
 :::
+:::{important}
+By default, the server applies `generation_config.json` from the Hugging Face model repository if it exists. This means the default values of certain sampling parameters can be overridden by those recommended by the model creator.
 
+To disable this behavior, please pass `--generation-config vllm` when launching the server.
+:::
 ## Supported APIs
 
 We currently support the following OpenAI APIs:

examples/offline_inference/basic/basic.py

@@ -18,7 +18,10 @@ llm = LLM(model="facebook/opt-125m")
 # that contain the prompt, generated text, and other information.
 outputs = llm.generate(prompts, sampling_params)
 # Print the outputs.
+print("\nGenerated Outputs:\n" + "-" * 60)
 for output in outputs:
     prompt = output.prompt
     generated_text = output.outputs[0].text
-    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
+    print(f"Prompt:    {prompt!r}")
+    print(f"Output:    {generated_text!r}")
+    print("-" * 60)

examples/offline_inference/basic/chat.py

@@ -27,12 +27,13 @@ def main(args: dict):
         sampling_params.top_k = top_k
 
     def print_outputs(outputs):
+        print("\nGenerated Outputs:\n" + "-" * 80)
         for output in outputs:
             prompt = output.prompt
             generated_text = output.outputs[0].text
-            print(f"Prompt: {prompt!r}")
+            print(f"Prompt: {prompt!r}\n")
             print(f"Generated text: {generated_text!r}")
-        print("-" * 80)
+            print("-" * 80)
 
     print("=" * 80)
 

examples/offline_inference/basic/classify.py

@@ -23,12 +23,14 @@ def main(args: Namespace):
     outputs = model.classify(prompts)
 
     # Print the outputs.
+    print("\nGenerated Outputs:\n" + "-" * 60)
     for prompt, output in zip(prompts, outputs):
         probs = output.outputs.probs
         probs_trimmed = ((str(probs[:16])[:-1] +
                           ", ...]") if len(probs) > 16 else probs)
-        print(f"Prompt: {prompt!r} | "
+        print(f"Prompt: {prompt!r} \n"
               f"Class Probabilities: {probs_trimmed} (size={len(probs)})")
+        print("-" * 60)
 
 
 if __name__ == "__main__":

examples/offline_inference/basic/embed.py

@@ -23,12 +23,14 @@ def main(args: Namespace):
     outputs = model.embed(prompts)
 
     # Print the outputs.
+    print("\nGenerated Outputs:\n" + "-" * 60)
     for prompt, output in zip(prompts, outputs):
         embeds = output.outputs.embedding
         embeds_trimmed = ((str(embeds[:16])[:-1] +
                            ", ...]") if len(embeds) > 16 else embeds)
-        print(f"Prompt: {prompt!r} | "
+        print(f"Prompt: {prompt!r} \n"
               f"Embeddings: {embeds_trimmed} (size={len(embeds)})")
+        print("-" * 60)
 
 
 if __name__ == "__main__":

examples/offline_inference/basic/score.py

@@ -22,9 +22,11 @@ def main(args: Namespace):
     outputs = model.score(text_1, texts_2)
 
     # Print the outputs.
+    print("\nGenerated Outputs:\n" + "-" * 60)
     for text_2, output in zip(texts_2, outputs):
         score = output.outputs.score
-        print(f"Pair: {[text_1, text_2]!r} | Score: {score}")
+        print(f"Pair: {[text_1, text_2]!r} \nScore: {score}")
+        print("-" * 60)
 
 
 if __name__ == "__main__":

examples/offline_inference/data_parallel.py

@@ -1,26 +1,49 @@
 # SPDX-License-Identifier: Apache-2.0
-# usage:
-# VLLM_USE_V1=1 python examples/offline_inference/data_parallel.py
-# we need to have a launcher to create multiple data parallel
-# ranks. And each rank will create a vLLM instance to process its own prompts.
+"""
+Usage:
+Single node:
+    python examples/offline_inference/data_parallel.py \
+            --model="ibm-research/PowerMoE-3b" \
+            --dp-size=2 \
+            --tp-size=2
+
+Multi-node:
+    Node 0 (assume the node has ip of 10.99.48.128):
+            python examples/offline_inference/data_parallel.py \
+                    --model="ibm-research/PowerMoE-3b" \
+                    --dp-size=2 \
+                    --tp-size=2 \
+                    --node-size=2 \
+                    --node-rank=0 \
+                    --master-addr=10.99.48.128 \
+                    --master-port=13345
+    Node 1:
+            python examples/offline_inference/data_parallel.py \
+                    --model="ibm-research/PowerMoE-3b" \
+                    --dp-size=2 \
+                    --tp-size=2 \
+                    --node-size=2 \
+                    --node-rank=1 \
+                    --master-addr=10.99.48.128 \
+                    --master-port=13345
+"""
 import os
+from time import sleep
 
 from vllm import LLM, SamplingParams
 from vllm.utils import get_open_port
 
-GPUs_per_dp_rank = 2
-DP_size = 2
 
-
-def main(dp_size, dp_rank, dp_master_ip, dp_master_port, GPUs_per_dp_rank):
-    os.environ["VLLM_DP_RANK"] = str(dp_rank)
+def main(model, dp_size, local_dp_rank, global_dp_rank, dp_master_ip,
+         dp_master_port, GPUs_per_dp_rank):
+    os.environ["VLLM_DP_RANK"] = str(global_dp_rank)
+    os.environ["VLLM_DP_RANK_LOCAL"] = str(local_dp_rank)
     os.environ["VLLM_DP_SIZE"] = str(dp_size)
     os.environ["VLLM_DP_MASTER_IP"] = dp_master_ip
     os.environ["VLLM_DP_MASTER_PORT"] = str(dp_master_port)
-    # set devices for each dp_rank
-    os.environ["CUDA_VISIBLE_DEVICES"] = ",".join(
-        str(i) for i in range(dp_rank * GPUs_per_dp_rank, (dp_rank + 1) *
-                              GPUs_per_dp_rank))
+
+    # CUDA_VISIBLE_DEVICES for each DP rank is set automatically inside the
+    # engine processes.
 
     # Sample prompts.
     prompts = [
@@ -28,20 +51,20 @@ def main(dp_size, dp_rank, dp_master_ip, dp_master_port, GPUs_per_dp_rank):
         "The president of the United States is",
         "The capital of France is",
         "The future of AI is",
-    ]
+    ] * 100
 
     # with DP, each rank should process different prompts.
     # usually all the DP ranks process a full dataset,
     # and each rank processes a different part of the dataset.
     promts_per_rank = len(prompts) // dp_size
-    start = dp_rank * promts_per_rank
+    start = global_dp_rank * promts_per_rank
     end = start + promts_per_rank
     prompts = prompts[start:end]
     if len(prompts) == 0:
         # if any rank has no prompts to process,
         # we need to set a placeholder prompt
         prompts = ["Placeholder"]
-    print(f"DP rank {dp_rank} needs to process {len(prompts)} prompts")
+    print(f"DP rank {global_dp_rank} needs to process {len(prompts)} prompts")
 
     # Create a sampling params object.
     # since we are doing data parallel, every rank can have different
@@ -49,37 +72,96 @@ def main(dp_size, dp_rank, dp_master_ip, dp_master_port, GPUs_per_dp_rank):
     # ranks for demonstration.
     sampling_params = SamplingParams(temperature=0.8,
                                      top_p=0.95,
-                                     max_tokens=16 * (dp_rank + 1))
+                                     max_tokens=[16, 20][global_dp_rank % 2])
 
     # Create an LLM.
-    llm = LLM(model="ibm-research/PowerMoE-3b",
+    llm = LLM(model=model,
               tensor_parallel_size=GPUs_per_dp_rank,
               enforce_eager=True,
               enable_expert_parallel=True)
     outputs = llm.generate(prompts, sampling_params)
     # Print the outputs.
-    for output in outputs:
+    for i, output in enumerate(outputs):
+        if i >= 5:
+            # print only 5 outputs
+            break
         prompt = output.prompt
         generated_text = output.outputs[0].text
-        print(f"DP rank {dp_rank}, Prompt: {prompt!r}, "
+        print(f"DP rank {global_dp_rank}, Prompt: {prompt!r}, "
               f"Generated text: {generated_text!r}")
 
+    # Give engines time to pause their processing loops before exiting.
+    sleep(1)
+
 
 if __name__ == "__main__":
+    import argparse
+    parser = argparse.ArgumentParser(description="Data Parallel Inference")
+    parser.add_argument("--model",
+                        type=str,
+                        default="ibm-research/PowerMoE-3b",
+                        help="Model name or path")
+    parser.add_argument("--dp-size",
+                        type=int,
+                        default=2,
+                        help="Data parallel size")
+    parser.add_argument("--tp-size",
+                        type=int,
+                        default=2,
+                        help="Tensor parallel size")
+    parser.add_argument("--node-size",
+                        type=int,
+                        default=1,
+                        help="Total number of nodes")
+    parser.add_argument("--node-rank",
+                        type=int,
+                        default=0,
+                        help="Rank of the current node")
+    parser.add_argument("--master-addr",
+                        type=str,
+                        default="",
+                        help="Master node IP address")
+    parser.add_argument("--master-port",
+                        type=int,
+                        default=0,
+                        help="Master node port")
+    args = parser.parse_args()
+
+    dp_size = args.dp_size
+    tp_size = args.tp_size
+    node_size = args.node_size
+    node_rank = args.node_rank
+
+    if node_size == 1:
+        dp_master_ip = "127.0.0.1"
+        dp_master_port = get_open_port()
+    else:
+        dp_master_ip = args.master_addr
+        dp_master_port = args.master_port
+
+    assert dp_size % node_size == 0, "dp_size should be divisible by node_size"
+    dp_per_node = dp_size // node_size
+
     from multiprocessing import Process
-    dp_master_ip = "127.0.0.1"
-    dp_master_port = get_open_port()
+
     procs = []
-    for i in range(DP_size):
+    for local_dp_rank, global_dp_rank in enumerate(
+            range(node_rank * dp_per_node, (node_rank + 1) * dp_per_node)):
         proc = Process(target=main,
-                       args=(DP_size, i, dp_master_ip, dp_master_port,
-                             GPUs_per_dp_rank))
+                       args=(args.model, dp_size, local_dp_rank,
+                             global_dp_rank, dp_master_ip, dp_master_port,
+                             tp_size))
         proc.start()
         procs.append(proc)
     exit_code = 0
     for proc in procs:
-        proc.join()
-        if proc.exitcode:
+        proc.join(timeout=300)
+        if proc.exitcode is None:
+            print(f"Killing process {proc.pid} that "
+                  f"didn't stop within 5 minutes.")
+            proc.kill()
+            exit_code = 1
+        elif proc.exitcode:
             exit_code = proc.exitcode
 
     exit(exit_code)

examples/offline_inference/mlpspeculator.py

@@ -50,7 +50,9 @@ if __name__ == "__main__":
     # Create an LLM with spec decoding
     llm = LLM(
         model="meta-llama/Llama-2-13b-chat-hf",
-        speculative_model="ibm-ai-platform/llama-13b-accelerator",
+        speculative_config={
+            "model": "ibm-ai-platform/llama-13b-accelerator",
+        },
     )
 
     print("With speculation")

examples/offline_inference/tpu.py

@@ -14,10 +14,7 @@ answers = [
 ]
 N = 1
 # Currently, top-p sampling is disabled. `top_p` should be 1.0.
-sampling_params = SamplingParams(temperature=0.7,
-                                 top_p=1.0,
-                                 n=N,
-                                 max_tokens=16)
+sampling_params = SamplingParams(temperature=0, top_p=1.0, n=N, max_tokens=16)
 
 # Set `enforce_eager=True` to avoid ahead-of-time compilation.
 # In real workloads, `enforace_eager` should be `False`.

examples/offline_inference/vision_language.py

@@ -361,6 +361,7 @@ def run_llava_next_video(questions: list[str],
     engine_args = EngineArgs(
         model="llava-hf/LLaVA-NeXT-Video-7B-hf",
         max_model_len=8192,
+        max_num_seqs=2,
         disable_mm_preprocessor_cache=args.disable_mm_preprocessor_cache,
     )
 

examples/online_serving/disaggregated_prefill_zmq.sh

@@ -1,123 +0,0 @@
-#!/bin/bash
-# This file demonstrates the example usage of disaggregated prefilling with ZMQ
-# We will launch 2 vllm instances (1 for prefill and 1 for decode),
-# and then transfer the KV cache between them.
-
-set -xe
-
-echo "🚧🚧 Warning: The usage of disaggregated prefill is experimental and subject to change 🚧🚧"
-sleep 1
-
-# Trap the SIGINT signal (triggered by Ctrl+C)
-trap 'cleanup' INT
-
-# Cleanup function
-cleanup() {
-    echo "Caught Ctrl+C, cleaning up..."
-    # Cleanup commands
-    pgrep python | xargs kill -9
-    pkill -f python
-    echo "Cleanup complete. Exiting."
-    exit 0
-}
-
-export VLLM_HOST_IP=$(hostname -I | awk '{print $1}')
-
-# a function that waits vLLM connect to start
-wait_for_server() {
-  local port=$1
-  timeout 1200 bash -c "
-    until curl -s localhost:${port}/v1/completions > /dev/null; do
-      sleep 1
-    done" && return 0 || return 1
-}
-
-
-# a function that waits vLLM disagg to start
-wait_for_disagg_server() {
-  local log_file=$1
-  timeout 1200 bash -c "
-    until grep -q 'PDWorker is ready' $log_file; do
-      sleep 1
-    done" && return 0 || return 1
-}
-
-
-# You can also adjust --kv-ip and --kv-port for distributed inference.
-MODEL=meta-llama/Llama-3.1-8B-Instruct
-CONTROLLER_ADDR=controller.ipc
-PREFILL_WORKER_ADDR=prefill.ipc
-DECODE_WORKER_ADDR=decode.ipc
-PORT=8001
-
-# prefilling instance, which is the KV producer
-CUDA_VISIBLE_DEVICES=0 python3 ../../vllm/entrypoints/disaggregated/worker.py \
-    --model $MODEL \
-    --controller-addr $CONTROLLER_ADDR \
-    --worker-addr $PREFILL_WORKER_ADDR \
-    --max-model-len 100 \
-    --gpu-memory-utilization 0.8 \
-    --kv-transfer-config \
-    '{"kv_connector":"PyNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2}' > vllm_disagg_prefill.log 2>&1 &
-
-# decoding instance, which is the KV consumer
-CUDA_VISIBLE_DEVICES=1 python3 ../../vllm/entrypoints/disaggregated/worker.py \
-    --model $MODEL \
-    --controller-addr $CONTROLLER_ADDR \
-    --worker-addr $DECODE_WORKER_ADDR \
-    --max-model-len 100 \
-    --gpu-memory-utilization 0.8 \
-    --kv-transfer-config \
-    '{"kv_connector":"PyNcclConnector","kv_role":"kv_consumer","kv_rank":1,"kv_parallel_size":2}' > vllm_disagg_decode.log 2>&1 &
-
-# launch a proxy server that opens the service at port 8000
-# the workflow of this proxy:
-# - Send req to prefill instance, wait until complete.
-# - Send req to decode instance, streaming tokens.
-python3 ../../vllm/entrypoints/disaggregated/api_server.py \
-    --port $PORT \
-    --model $MODEL \
-    --controller-addr $CONTROLLER_ADDR \
-    --prefill-addr $PREFILL_WORKER_ADDR \
-    --decode-addr $DECODE_WORKER_ADDR &
-
-# wait until prefill, decode instances and proxy are ready
-wait_for_server $PORT
-wait_for_disagg_server vllm_disagg_prefill.log
-wait_for_disagg_server vllm_disagg_decode.log 
-
-# serve two example requests
-output1=$(curl -X POST -s http://localhost:8001/v1/completions \
--H "Content-Type: application/json" \
--d '{
-"model": "meta-llama/Llama-3.1-8B-Instruct",
-"prompt": "San Francisco is a",
-"max_tokens": 10,
-"temperature": 0
-}')
-
-output2=$(curl -X POST -s http://localhost:8001/v1/completions \
--H "Content-Type: application/json" \
--d '{
-"model": "meta-llama/Llama-3.1-8B-Instruct",
-"prompt": "Santa Clara is a",
-"max_tokens": 10,
-"temperature": 0
-}')
-
-
-# Cleanup commands
-pgrep python | xargs kill -9
-pkill -f python
-
-echo ""
-
-sleep 1
-
-# Print the outputs of the curl requests
-echo ""
-echo "Output of first request: $output1"
-echo "Output of second request: $output2"
-
-echo "🎉🎉 Successfully finished 2 test requests! 🎉🎉"
-echo ""

examples/online_serving/openai_chat_completion_tool_calls_with_reasoning.py

@@ -0,0 +1,177 @@
+# SPDX-License-Identifier: Apache-2.0
+"""
+An example demonstrates how to use tool calling with reasoning models 
+like QwQ-32B. The reasoning_content will not be parsed by the tool 
+calling process; only the final output will be parsed.
+
+To run this example, you need to start the vLLM server with both 
+the reasoning parser and tool calling enabled.
+
+```bash
+vllm serve Qwen/QwQ-32B \
+     --enable-reasoning --reasoning-parser deepseek_r1 \
+     --enable-auto-tool-choice --tool-call-parser hermes
+     
+```
+
+"""
+
+from openai import OpenAI
+
+
+# Now, simulate a tool call
+def get_current_weather(city: str, state: str, unit: 'str'):
+    return ("The weather in Dallas, Texas is 85 degrees fahrenheit. It is "
+            "partly cloudly, with highs in the 90's.")
+
+
+available_tools = {"get_current_weather": get_current_weather}
+
+# Modify OpenAI's API key and API base to use vLLM's API server.
+openai_api_key = "EMPTY"
+openai_api_base = "http://localhost:8000/v1"
+
+client = OpenAI(
+    api_key=openai_api_key,
+    base_url=openai_api_base,
+)
+
+models = client.models.list()
+model = models.data[0].id
+
+tools = [{
+    "type": "function",
+    "function": {
+        "name": "get_current_weather",
+        "description": "Get the current weather in a given location",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "city": {
+                    "type":
+                    "string",
+                    "description":
+                    "The city to find the weather for, e.g. 'San Francisco'"
+                },
+                "state": {
+                    "type":
+                    "string",
+                    "description":
+                    "the two-letter abbreviation for the state that the city is"
+                    " in, e.g. 'CA' which would mean 'California'"
+                },
+                "unit": {
+                    "type": "string",
+                    "description": "The unit to fetch the temperature in",
+                    "enum": ["celsius", "fahrenheit"]
+                }
+            },
+            "required": ["city", "state", "unit"]
+        }
+    }
+}]
+messages = [{
+    "role": "user",
+    "content": "Hi! How are you doing today?"
+}, {
+    "role": "assistant",
+    "content": "I'm doing well! How can I help you?"
+}, {
+    "role":
+    "user",
+    "content":
+    "Can you tell me what the temperate will be in Dallas, in fahrenheit?"
+}]
+
+
+def extract_reasoning_and_calls(chunks: list):
+    reasoning_content = ""
+    tool_call_idx = -1
+    arguments = []
+    function_names = []
+    for chunk in chunks:
+        if chunk.choices[0].delta.tool_calls:
+            tool_call = chunk.choices[0].delta.tool_calls[0]
+            if tool_call.index != tool_call_idx:
+                tool_call_idx = chunk.choices[0].delta.tool_calls[0].index
+                arguments.append("")
+                function_names.append("")
+
+            if tool_call.function:
+                if tool_call.function.name:
+                    function_names[tool_call_idx] = tool_call.function.name
+
+                if tool_call.function.arguments:
+                    arguments[tool_call_idx] += tool_call.function.arguments
+        else:
+            if hasattr(chunk.choices[0].delta, "reasoning_content"):
+                reasoning_content += chunk.choices[0].delta.reasoning_content
+    return reasoning_content, arguments, function_names
+
+
+print("---------Full Generate With Automatic Function Calling-------------")
+tool_calls = client.chat.completions.create(messages=messages,
+                                            model=model,
+                                            tools=tools)
+print(f"reasoning_content: {tool_calls.choices[0].message.reasoning_content}")
+print(f"function name: "
+      f"{tool_calls.choices[0].message.tool_calls[0].function.name}")
+print(f"function arguments: "
+      f"{tool_calls.choices[0].message.tool_calls[0].function.arguments}")
+
+print("----------Stream Generate With Automatic Function Calling-----------")
+tool_calls_stream = client.chat.completions.create(messages=messages,
+                                                   model=model,
+                                                   tools=tools,
+                                                   stream=True)
+chunks = []
+for chunk in tool_calls_stream:
+    chunks.append(chunk)
+
+reasoning_content, arguments, function_names = extract_reasoning_and_calls(
+    chunks)
+
+print(f"reasoning_content: {reasoning_content}")
+print(f"function name: {function_names[0]}")
+print(f"function arguments: {arguments[0]}")
+
+print("----------Full Generate With Named Function Calling-----------------")
+tool_calls = client.chat.completions.create(messages=messages,
+                                            model=model,
+                                            tools=tools,
+                                            tool_choice={
+                                                "type": "function",
+                                                "function": {
+                                                    "name":
+                                                    "get_current_weather"
+                                                }
+                                            })
+
+tool_call = tool_calls.choices[0].message.tool_calls[0].function
+print(f"reasoning_content: {tool_calls.choices[0].message.reasoning_content}")
+print(f"function name: {tool_call.name}")
+print(f"function arguments: {tool_call.arguments}")
+print("----------Stream Generate With Named Function Calling--------------")
+
+tool_calls_stream = client.chat.completions.create(
+    messages=messages,
+    model=model,
+    tools=tools,
+    tool_choice={
+        "type": "function",
+        "function": {
+            "name": "get_current_weather"
+        }
+    },
+    stream=True)
+
+chunks = []
+for chunk in tool_calls_stream:
+    chunks.append(chunk)
+
+reasoning_content, arguments, function_names = extract_reasoning_and_calls(
+    chunks)
+print(f"reasoning_content: {reasoning_content}")
+print(f"function name: {function_names[0]}")
+print(f"function arguments: {arguments[0]}")
+print("\n\n")

examples/online_serving/openai_chat_completion_with_reasoning.py

@@ -31,6 +31,7 @@ model = models.data[0].id
 
 # Round 1
 messages = [{"role": "user", "content": "9.11 and 9.8, which is greater?"}]
+# For granite, add: `extra_body={"chat_template_kwargs": {"thinking": True}}`
 response = client.chat.completions.create(model=model, messages=messages)
 
 reasoning_content = response.choices[0].message.reasoning_content

examples/online_serving/openai_chat_completion_with_reasoning_streaming.py

@@ -38,6 +38,7 @@ models = client.models.list()
 model = models.data[0].id
 
 messages = [{"role": "user", "content": "9.11 and 9.8, which is greater?"}]
+# For granite, add: `extra_body={"chat_template_kwargs": {"thinking": True}}`
 stream = client.chat.completions.create(model=model,
                                         messages=messages,
                                         stream=True)

pyproject.toml

@@ -86,6 +86,7 @@ exclude = [
 "vllm/triton_utils/**/*.py" = ["UP006", "UP035"]
 "vllm/vllm_flash_attn/**/*.py" = ["UP006", "UP035"]
 "vllm/worker/**/*.py" = ["UP006", "UP035"]
+"vllm/utils.py" = ["UP006", "UP035"]
 
 [tool.ruff.lint]
 select = [

requirements/common.txt

@@ -18,7 +18,7 @@ pillow  # Required for image processing
 prometheus-fastapi-instrumentator >= 7.0.0
 tiktoken >= 0.6.0  # Required for DBRX tokenizer
 lm-format-enforcer >= 0.10.11, < 0.11
-llguidance >= 0.7.2, < 0.8.0; platform_machine == "x86_64" or platform_machine == "arm64" or platform_machine == "aarch64"
+llguidance >= 0.7.9, < 0.8.0; platform_machine == "x86_64" or platform_machine == "arm64" or platform_machine == "aarch64"
 outlines == 0.1.11
 lark == 1.2.2
 xgrammar == 0.1.16; platform_machine == "x86_64" or platform_machine == "aarch64"

requirements/cpu.txt

@@ -4,14 +4,14 @@
 # Dependencies for CPUs
 torch==2.6.0+cpu; platform_machine == "x86_64"
 torch==2.6.0; platform_system == "Darwin"
-torch==2.5.1; platform_machine == "ppc64le" or platform_machine == "aarch64"
+torch==2.6.0; platform_machine == "ppc64le" or platform_machine == "aarch64"
 torch==2.7.0.dev20250304; platform_machine == "s390x"
 
 # required for the image processor of minicpm-o-2_6, this must be updated alongside torch
 torchaudio; platform_machine != "ppc64le" and platform_machine != "s390x"
-torchaudio==2.5.1; platform_machine == "ppc64le"
+torchaudio==2.6.0; platform_machine == "ppc64le"
 
 # required for the image processor of phi3v, this must be updated alongside torch
 torchvision; platform_machine != "ppc64le"  and platform_machine != "s390x"
-torchvision==0.20.1; platform_machine == "ppc64le"
+torchvision==0.21.0; platform_machine == "ppc64le"
 datasets # for benchmark scripts

requirements/cuda.txt

@@ -4,7 +4,7 @@
 numba == 0.60.0 # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
 
 # Dependencies for NVIDIA GPUs
-ray[cgraph]>=2.43.0 # Ray Compiled Graph, required for pipeline parallelism in V1.
+ray[cgraph]>=2.43.0, !=2.44.* # Ray Compiled Graph, required for pipeline parallelism in V1.
 torch==2.6.0
 torchaudio==2.6.0
 # These must be updated alongside torch

requirements/test.in

@@ -17,7 +17,7 @@ vector_quantize_pytorch # required for minicpmo_26 test
 vocos # required for minicpmo_26 test
 peft
 pqdm
-ray[cgraph]>=2.43.0 # Ray Compiled Graph, required by pipeline parallelism tests
+ray[cgraph]>=2.43.0, !=2.44.* # Ray Compiled Graph, required by pipeline parallelism tests
 sentence-transformers # required for embedding tests
 soundfile # required for audio tests
 jiwer # required for audio tests
@@ -41,3 +41,4 @@ tritonclient==2.51.0
 numpy < 2.0.0
 runai-model-streamer==0.11.0
 runai-model-streamer-s3==0.11.0
+fastsafetensors>=0.1.10

requirements/test.txt

@@ -67,6 +67,7 @@ click==8.1.7
     #   jiwer
     #   nltk
     #   ray
+    #   typer
 colorama==0.4.6
     # via
     #   awscli
@@ -122,6 +123,8 @@ fastparquet==2024.11.0
     # via genai-perf
 fastrlock==0.8.2
     # via cupy-cuda12x
+fastsafetensors==0.1.10
+    # via -r requirements/test.in
 filelock==3.16.1
     # via
     #   datasets
@@ -505,7 +508,9 @@ requests==2.32.3
 responses==0.25.3
     # via genai-perf
 rich==13.9.4
-    # via genai-perf
+    # via
+    #   genai-perf
+    #   typer
 rouge-score==0.1.2
     # via lm-eval
 rpds-py==0.20.1
@@ -550,6 +555,8 @@ setuptools==75.8.0
     # via
     #   pytablewriter
     #   torch
+shellingham==1.5.4
+    # via typer
 six==1.16.0
     # via
     #   python-dateutil
@@ -600,6 +607,7 @@ torch==2.6.0
     #   accelerate
     #   bitsandbytes
     #   encodec
+    #   fastsafetensors
     #   lm-eval
     #   peft
     #   runai-model-streamer
@@ -654,6 +662,8 @@ typepy==1.3.2
     #   dataproperty
     #   pytablewriter
     #   tabledata
+typer==0.15.2
+    # via fastsafetensors
 typing-extensions==4.12.2
     # via
     #   huggingface-hub
@@ -663,6 +673,7 @@ typing-extensions==4.12.2
     #   pydantic
     #   pydantic-core
     #   torch
+    #   typer
 tzdata==2024.2
     # via pandas
 urllib3==2.2.3

setup.py

@@ -680,6 +680,7 @@ setup(
     install_requires=get_requirements(),
     extras_require={
         "tensorizer": ["tensorizer>=2.9.0"],
+        "fastsafetensors": ["fastsafetensors >= 0.1.10"],
         "runai": ["runai-model-streamer", "runai-model-streamer-s3", "boto3"],
         "audio": ["librosa", "soundfile"],  # Required for audio processing
         "video": ["decord"]  # Required for video processing

tests/build_cython.py

@@ -0,0 +1,38 @@
+# SPDX-License-Identifier: Apache-2.0
+import Cython.Compiler.Options
+from Cython.Build import cythonize
+from setuptools import setup
+
+Cython.Compiler.Options.annotate = True
+
+infiles = []
+
+infiles += [
+    "vllm/engine/llm_engine.py",
+    "vllm/transformers_utils/detokenizer.py",
+    "vllm/engine/output_processor/single_step.py",
+    "vllm/outputs.py",
+    "vllm/engine/output_processor/stop_checker.py",
+]
+
+infiles += [
+    "vllm/core/scheduler.py",
+    "vllm/sequence.py",
+    "vllm/core/block_manager.py",
+]
+
+infiles += [
+    "vllm/model_executor/layers/sampler.py",
+    "vllm/sampling_params.py",
+    "vllm/utils.py",
+]
+
+setup(ext_modules=cythonize(infiles,
+                            annotate=False,
+                            force=True,
+                            compiler_directives={
+                                'language_level': "3",
+                                'infer_types': True
+                            }))
+
+# example usage: python3 build_cython.py build_ext --inplace

tests/compile/piecewise/test_toy_llama.py

@@ -63,7 +63,8 @@ class LlamaConfig:
             factors.append((k, v))
         factors.sort()
         import hashlib
-        return hashlib.md5(str(factors).encode()).hexdigest()
+        return hashlib.md5(str(factors).encode(),
+                           usedforsecurity=False).hexdigest()
 
     def __post_init__(self):
         assert self.mlp_size >= self.hidden_size

tests/compile/test_pass_manager.py

@@ -1,6 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-
-import pickle
+import copy
 
 import pytest
 import torch
@@ -10,32 +9,63 @@ from vllm.compilation.pass_manager import PostGradPassManager
 from vllm.config import CompilationConfig
 
 
+# dummy custom pass that doesn't inherit
 def simple_callable(graph: torch.fx.Graph):
     pass
 
 
-callable_uuid = CallableInductorPass(simple_callable,
-                                     InductorPass.hash_source(__file__))
-
-
-@pytest.mark.parametrize(
-    "works, callable",
-    [
-        (False, simple_callable),
-        (True, callable_uuid),
-        (True, CallableInductorPass(simple_callable)),
-    ],
-)
-def test_pass_manager(works: bool, callable):
+# Should fail to add directly to the pass manager
+def test_bad_callable():
     config = CompilationConfig().pass_config
 
     pass_manager = PostGradPassManager()
     pass_manager.configure(config)
 
-    # Try to add the callable to the pass manager
-    if works:
-        pass_manager.add(callable)
-        pickle.dumps(pass_manager)
-    else:
-        with pytest.raises(AssertionError):
-            pass_manager.add(callable)
+    with pytest.raises(AssertionError):
+        pass_manager.add(simple_callable)  # noqa, type wrong on purpose
+
+
+# Pass that inherits from InductorPass
+class ProperPass(InductorPass):
+
+    def __call__(self, graph: torch.fx.graph.Graph) -> None:
+        pass
+
+
+@pytest.mark.parametrize(
+    "callable",
+    [
+        ProperPass(),
+        # Can also wrap callables in CallableInductorPass for compliance
+        CallableInductorPass(simple_callable),
+        CallableInductorPass(simple_callable,
+                             InductorPass.hash_source(__file__))
+    ],
+)
+def test_pass_manager_uuid(callable):
+    config = CompilationConfig().pass_config
+
+    pass_manager = PostGradPassManager()
+    pass_manager.configure(config)
+
+    # Check that UUID is different if the same pass is added 2x
+    pass_manager.add(callable)
+    uuid1 = pass_manager.uuid()
+    pass_manager.add(callable)
+    uuid2 = pass_manager.uuid()
+    assert uuid1 != uuid2
+
+    # UUID should be the same as the original one,
+    # as we constructed in the same way.
+    pass_manager2 = PostGradPassManager()
+    pass_manager2.configure(config)
+    pass_manager2.add(callable)
+    assert uuid1 == pass_manager2.uuid()
+
+    # UUID should be different due to config change
+    config2 = copy.deepcopy(config)
+    config2.enable_fusion = not config2.enable_fusion
+    pass_manager3 = PostGradPassManager()
+    pass_manager3.configure(config2)
+    pass_manager3.add(callable)
+    assert uuid1 != pass_manager3.uuid()

tests/distributed/test_pipeline_parallel.py

@@ -175,6 +175,8 @@ TEXT_GENERATION_MODELS = {
     "inceptionai/jais-13b-chat": PPTestSettings.fast(),
     "ai21labs/Jamba-tiny-dev": PPTestSettings.fast(),
     "meta-llama/Llama-3.2-1B-Instruct": PPTestSettings.detailed(),
+    # Tests TransformersForCausalLM
+    "ArthurZ/Ilama-3.2-1B": PPTestSettings.fast(),
     "openbmb/MiniCPM-2B-sft-bf16": PPTestSettings.fast(),
     "openbmb/MiniCPM3-4B": PPTestSettings.fast(),
     # Uses Llama
@@ -243,6 +245,7 @@ TEST_MODELS = [
     # [LANGUAGE GENERATION]
     "microsoft/Phi-3.5-MoE-instruct",
     "meta-llama/Llama-3.2-1B-Instruct",
+    # "ArthurZ/Ilama-3.2-1B", NOTE: Uncomment after #13905
     "ibm/PowerLM-3b",
     # [LANGUAGE EMBEDDING]
     "intfloat/e5-mistral-7b-instruct",

tests/entrypoints/llm/test_collective_rpc.py

@@ -21,18 +21,9 @@ def test_collective_rpc(tp_size, backend):
     def echo_rank(self):
         return self.rank
 
-    from vllm.worker.worker import Worker
-
-    class MyWorker(Worker):
-
-        def echo_rank(self):
-            return self.rank
-
     llm = LLM(model="meta-llama/Llama-3.2-1B-Instruct",
               enforce_eager=True,
               load_format="dummy",
               tensor_parallel_size=tp_size,
-              distributed_executor_backend=backend,
-              worker_cls=MyWorker)
-    for method in ["echo_rank", echo_rank]:
-        assert llm.collective_rpc(method) == list(range(tp_size))
+              distributed_executor_backend=backend)
+    assert llm.collective_rpc(echo_rank) == list(range(tp_size))

tests/entrypoints/openai/reasoning_parsers/test_granite_reasoning_parser.py

@@ -0,0 +1,349 @@
+# SPDX-License-Identifier: Apache-2.0
+import pytest
+from transformers import AutoTokenizer
+
+from tests.entrypoints.openai.reasoning_parsers.utils import (
+    DeltaMessage, run_reasoning_extraction)
+from vllm.entrypoints.openai.reasoning_parsers import (ReasoningParser,
+                                                       ReasoningParserManager)
+
+parser_name = "granite"
+START_REASONING = "Here is my thought process:"
+START_RESPONSE = "Here is my response:"
+
+SIMPLE_REASONING = {
+    "output":
+    f"{START_REASONING}This is a reasoning section{START_RESPONSE}This is the rest",  #noqa: E501
+    "reasoning_content": "This is a reasoning section",
+    "content": "This is the rest",
+}
+COMPLETE_REASONING = {
+    "output": f"{START_REASONING}This is a reasoning section{START_RESPONSE}",
+    "reasoning_content": "This is a reasoning section",
+    "content": None,
+}
+NO_REASONING = {
+    "output": "This is content",
+    "reasoning_content": None,
+    "content": "This is content",
+}
+MULTIPLE_LINES = {
+    "output":
+    f"{START_REASONING}This\nThat{START_RESPONSE}This is the rest\nThat",
+    "reasoning_content": "This\nThat",
+    "content": "This is the rest\nThat",
+}
+REASONING_WITH_THINK = {
+    "output":
+    f"{START_REASONING}This is a reasoning section{START_RESPONSE}This is the rest",  #noqa: E501
+    "reasoning_content": "This is a reasoning section",
+    "content": "This is the rest",
+}
+COMPLETE_REASONING_WITH_THINK = {
+    "output": f"{START_REASONING}This is a reasoning section{START_RESPONSE}",
+    "reasoning_content": "This is a reasoning section",
+    "content": None,
+}
+MULTIPLE_LINES_WITH_THINK = {
+    "output":
+    f"{START_REASONING}This\nThat{START_RESPONSE}This is the rest\nThat",
+    "reasoning_content": "This\nThat",
+    "content": "This is the rest\nThat",
+}
+
+TEST_CASES = [
+    pytest.param(
+        False,
+        SIMPLE_REASONING,
+        id="simple_reasoning",
+    ),
+    pytest.param(
+        False,
+        COMPLETE_REASONING,
+        id="complete_reasoning",
+    ),
+    pytest.param(
+        False,
+        NO_REASONING,
+        id="no_reasoning",
+    ),
+    pytest.param(
+        False,
+        MULTIPLE_LINES,
+        id="multiple_lines",
+    ),
+    pytest.param(
+        False,
+        REASONING_WITH_THINK,
+        id="reasoning_with_think",
+    ),
+    pytest.param(
+        False,
+        COMPLETE_REASONING_WITH_THINK,
+        id="complete_reasoning_with_think",
+    ),
+    pytest.param(
+        False,
+        MULTIPLE_LINES_WITH_THINK,
+        id="multiple_lines_with_think",
+    ),
+    pytest.param(
+        True,
+        SIMPLE_REASONING,
+        id="simple_reasoning_streaming",
+    ),
+    pytest.param(
+        True,
+        COMPLETE_REASONING,
+        id="complete_reasoning_streaming",
+    ),
+    pytest.param(
+        True,
+        NO_REASONING,
+        id="no_reasoning_streaming",
+    ),
+    pytest.param(
+        True,
+        MULTIPLE_LINES,
+        id="multiple_lines_streaming",
+    ),
+    pytest.param(
+        True,
+        REASONING_WITH_THINK,
+        id="reasoning_with_think_streaming",
+    ),
+    pytest.param(
+        True,
+        COMPLETE_REASONING_WITH_THINK,
+        id="complete_reasoning_with_think_streaming",
+    ),
+    pytest.param(
+        True,
+        MULTIPLE_LINES_WITH_THINK,
+        id="multiple_lines_with_think_streaming",
+    ),
+]
+
+# Global tokenizer initialization to avoid repeated loading
+tokenizer = AutoTokenizer.from_pretrained("facebook/opt-125m")
+
+
+@pytest.mark.parametrize("streaming, param_dict", TEST_CASES)
+def test_reasoning(
+    streaming: bool,
+    param_dict: dict,
+):
+    output = tokenizer.tokenize(param_dict["output"])
+    # decode everything to tokens
+    output_tokens: list[str] = [
+        tokenizer.convert_tokens_to_string([token]) for token in output
+    ]
+    parser: ReasoningParser = ReasoningParserManager.get_reasoning_parser(
+        parser_name)(tokenizer)
+
+    reasoning, content = run_reasoning_extraction(parser,
+                                                  output_tokens,
+                                                  streaming=streaming)
+
+    assert reasoning == param_dict["reasoning_content"]
+    assert content == param_dict["content"]
+
+
+# Additional tests for verifying the correctness of granite streaming; this
+# is complicated because granite uses multiple tokens to indicate when thinking
+# is starting / when it's starting its response, so skipping special tokens
+# is awkward.
+
+### Handling the start of reasoning
+STREAMING_1 = {
+    "previous_text": None,
+    "current_text": "Here",
+    "delta_text": "Here",
+    "reasoning_content": None,
+    "content": None,
+}
+# When we fail, we should give what was previously being silenced first
+STREAMING_2 = {
+    "previous_text": "Here is my thought",
+    "current_text": "Here is my thought failure",
+    "delta_text": " failure",
+    "reasoning_content": None,
+    "content": "Here is my thought failure",
+}
+# But then after the first one, we should only add the delta text to content
+STREAMING_3 = {
+    "previous_text": "Here wrong",
+    "current_text": " words",
+    "delta_text": " Here wrong words",
+    "reasoning_content": None,
+    "content": " words",
+}
+# But then after the first one, we should only add the delta text to content
+STREAMING_4 = {
+    "previous_text": "Here is my thought",
+    "current_text": "Here is my thought process:",
+    "delta_text": " process:",
+    "reasoning_content": None,
+    "content": None,
+}
+# Reasoning started successfully; parse reasoning content
+STREAMING_5 = {
+    "previous_text": "Here is my thought process:",
+    "current_text": "Here is my thought process: foo",
+    "delta_text": " foo",
+    "reasoning_content": " foo",
+    "content": None,
+}
+# Response special sequence has started, but not finished.
+STREAMING_6 = {
+    "previous_text": "Here is my thought process: foo",
+    "current_text": "Here is my thought process: foo Here is",
+    "delta_text": " Here is",
+    "reasoning_content": " ",
+    "content": None,
+}
+# Response special sequence started, but was broken; the reasoning
+# content should be the content that was previously unused.
+STREAMING_7 = {
+    "previous_text": "Here is my thought process: foo Here is",
+    "current_text": "Here is my thought process: foo Here is Here",
+    "delta_text": " Here",
+    "reasoning_content": "Here is ",
+    "content": None,
+}
+# Response special sequence is ongoing
+STREAMING_8 = {
+    "previous_text": "Here is my thought process: foo Here is my response:",
+    "current_text": "Here is my thought process: foo Here is my response: bar",
+    "delta_text": " bar",
+    "reasoning_content": None,
+    "content": " bar",
+}
+# The delta text has everything; we should be able to correctly parse both
+STREAMING_9 = {
+    "previous_text": None,
+    "current_text": "Here is my thought process: foo Here is my response: bar",
+    "delta_text": "Here is my thought process: foo Here is my response: bar",
+    "reasoning_content": " foo ",
+    "content": " bar",
+}
+## The Response is ongoing, and the delta mixes reasoning content / content
+STREAMING_10 = {
+    "previous_text": "Here is my thought process: foo",
+    "current_text":
+    "Here is my thought process: foo bar Here is my response: baz",
+    "delta_text": " bar Here is my response: baz",
+    "reasoning_content": " bar ",
+    "content": " baz",
+}
+# The delta text starts a new substring that might be a response special seq
+STREAMING_11 = {
+    "previous_text":
+    "Here is my thought process: This is a reasoning section ",
+    "current_text":
+    "Here is my thought process: This is a reasoning section Here",
+    "delta_text": "Here",
+    "reasoning_content": None,
+    "content": None,
+}
+# The delta text is finishing the response special seq
+STREAMING_12 = {
+    "previous_text": "Here is my thought process: foo Here is my response",
+    "current_text": "Here is my thought process: foo Here is my response:",
+    "delta_text": ":",
+    "reasoning_content": None,
+    "content": None,
+}
+STREAMING_13 = {
+    "previous_text": "Here is my thought process: foo Here",
+    "current_text": "Here is my thought process: foo Here was",
+    "delta_text": " was",
+    "reasoning_content": "Here was",
+    "content": None,
+}
+
+STREAMING_SUBCASES = [
+    pytest.param(
+        STREAMING_1,
+        id="Starting reasoning special sequence",
+    ),
+    pytest.param(
+        STREAMING_2,
+        id="Unexpected start reasoning sequence",
+    ),
+    pytest.param(
+        STREAMING_3,
+        id="Continuing unexpected start reasoning sequence",
+    ),
+    pytest.param(
+        STREAMING_4,
+        id="Only start reasoning sequence and nothing else",
+    ),
+    pytest.param(
+        STREAMING_5,
+        id="Reasoning content has started",
+    ),
+    pytest.param(
+        STREAMING_6,
+        id="Response special sequence has started",
+    ),
+    pytest.param(
+        STREAMING_7,
+        id="Response special sequence reset",
+    ),
+    pytest.param(
+        STREAMING_8,
+        id="Response text has started",
+    ),
+    pytest.param(
+        STREAMING_9,
+        id="Delta contains everything",
+    ),
+    pytest.param(
+        STREAMING_10,
+        id="Delta contains some reasoning and response",
+    ),
+    pytest.param(
+        STREAMING_11,
+        id="Delta starts response sequence",
+    ),
+    pytest.param(
+        STREAMING_12,
+        id="Delta finishes response sequence",
+    ),
+    pytest.param(
+        STREAMING_13,
+        id="Delta breaks potential responise sequence",
+    ),
+]
+
+
+@pytest.mark.parametrize("param_dict", STREAMING_SUBCASES)
+def test_streaming_subcases(param_dict):
+    # Get all of the token IDs
+    previous_token_ids = tokenizer.encode(
+        param_dict["previous_text"]
+    ) if param_dict["previous_text"] is not None else []
+    current_token_ids = tokenizer.encode(param_dict["current_text"])
+    delta_token_ids = tokenizer.encode(param_dict["delta_text"])
+
+    parser: ReasoningParser = ReasoningParserManager.get_reasoning_parser(
+        parser_name)(tokenizer)
+
+    response = parser.extract_reasoning_content_streaming(
+        previous_text=param_dict["previous_text"],
+        current_text=param_dict["current_text"],
+        delta_text=param_dict["delta_text"],
+        previous_token_ids=previous_token_ids,
+        current_token_ids=current_token_ids,
+        delta_token_ids=delta_token_ids,
+    )
+    # Streaming currently expects at least one of reasoning content / content,
+    # so the response should return None in that case.
+    if param_dict["reasoning_content"] is None and param_dict[
+            "content"] is None:
+        assert response is None
+    else:
+        assert isinstance(response, DeltaMessage)
+        assert param_dict["reasoning_content"] == response.reasoning_content
+        assert param_dict["content"] == response.content

tests/entrypoints/openai/test_chat_template.py

@@ -107,8 +107,10 @@ def test_get_gen_prompt(model, template, add_generation_prompt,
     # Call the function and get the result
     result = apply_hf_chat_template(
         tokenizer,
+        trust_remote_code=True,
         conversation=mock_request.messages,
         chat_template=mock_request.chat_template or template_content,
+        tools=None,
         add_generation_prompt=mock_request.add_generation_prompt,
         continue_final_message=mock_request.continue_final_message,
     )

tests/entrypoints/openai/test_chat_with_tool_reasoning.py

@@ -0,0 +1,145 @@
+# SPDX-License-Identifier: Apache-2.0
+
+import openai  # use the official client for correctness check
+import pytest
+import pytest_asyncio
+
+from ...utils import RemoteOpenAIServer
+
+# a reasoning and tool calling model
+MODEL_NAME = "Qwen/QwQ-32B"
+
+
+@pytest.fixture(scope="module")
+def server():  # noqa: F811
+    args = [
+        "--max-model-len", "8192", "--enforce-eager", "--enable-reasoning",
+        "--reasoning-parser", "deepseek_r1", "--enable-auto-tool-choice",
+        "--tool-call-parser", "hermes"
+    ]
+
+    with RemoteOpenAIServer(MODEL_NAME, args) as remote_server:
+        yield remote_server
+
+
+@pytest_asyncio.fixture
+async def client(server):
+    async with server.get_async_client() as async_client:
+        yield async_client
+
+
+TOOLS = [{
+    "type": "function",
+    "function": {
+        "name": "get_current_weather",
+        "description": "Get the current weather in a given location",
+        "parameters": {
+            "type": "object",
+            "properties": {
+                "city": {
+                    "type":
+                    "string",
+                    "description":
+                    "The city to find the weather for, e.g. 'San Francisco'"
+                },
+                "state": {
+                    "type":
+                    "string",
+                    "description":
+                    "the two-letter abbreviation for the state that the city is"
+                    " in, e.g. 'CA' which would mean 'California'"
+                },
+                "unit": {
+                    "type": "string",
+                    "description": "The unit to fetch the temperature in",
+                    "enum": ["celsius", "fahrenheit"]
+                }
+            },
+            "required": ["city", "state", "unit"]
+        }
+    }
+}]
+
+MESSAGES = [{
+    "role": "user",
+    "content": "Hi! How are you doing today?"
+}, {
+    "role": "assistant",
+    "content": "I'm doing well! How can I help you?"
+}, {
+    "role":
+    "user",
+    "content":
+    "Can you tell me what the temperate will be in Dallas, in fahrenheit?"
+}]
+
+FUNC_NAME = "get_current_weather"
+FUNC_ARGS = """{"city": "Dallas", "state": "TX", "unit": "fahrenheit"}"""
+
+
+def extract_reasoning_and_calls(chunks: list):
+    reasoning_content = ""
+    tool_call_idx = -1
+    arguments = []
+    function_names = []
+    for chunk in chunks:
+        if chunk.choices[0].delta.tool_calls:
+            tool_call = chunk.choices[0].delta.tool_calls[0]
+            if tool_call.index != tool_call_idx:
+                tool_call_idx = chunk.choices[0].delta.tool_calls[0].index
+                arguments.append("")
+                function_names.append("")
+
+            if tool_call.function:
+                if tool_call.function.name:
+                    function_names[tool_call_idx] = tool_call.function.name
+
+                if tool_call.function.arguments:
+                    arguments[tool_call_idx] += tool_call.function.arguments
+        else:
+            if hasattr(chunk.choices[0].delta, "reasoning_content"):
+                reasoning_content += chunk.choices[0].delta.reasoning_content
+    return reasoning_content, arguments, function_names
+
+
+# test streaming
+@pytest.mark.asyncio
+async def test_chat_streaming_of_tool_and_reasoning(
+        client: openai.AsyncOpenAI):
+
+    stream = await client.chat.completions.create(
+        model=MODEL_NAME,
+        messages=MESSAGES,
+        tools=TOOLS,
+        temperature=0.0,
+        stream=True,
+    )
+
+    chunks = []
+    async for chunk in stream:
+        chunks.append(chunk)
+
+    reasoning_content, arguments, function_names = extract_reasoning_and_calls(
+        chunks)
+    assert len(reasoning_content) > 0
+    assert len(function_names) > 0 and function_names[0] == FUNC_NAME
+    assert len(arguments) > 0 and arguments[0] == FUNC_ARGS
+
+
+# test full generate
+@pytest.mark.asyncio
+async def test_chat_full_of_tool_and_reasoning(client: openai.AsyncOpenAI):
+
+    tool_calls = await client.chat.completions.create(
+        model=MODEL_NAME,
+        messages=MESSAGES,
+        tools=TOOLS,
+        temperature=0.0,
+        stream=False,
+    )
+
+    assert len(tool_calls.choices[0].message.reasoning_content) > 0
+    assert tool_calls.choices[0].message.tool_calls[0].function.name \
+          == FUNC_NAME
+    assert tool_calls.choices[0].message.tool_calls[0].function.arguments \
+          == FUNC_ARGS

tests/entrypoints/openai/test_video.py

@@ -87,7 +87,7 @@ async def test_single_chat_session_video(client: openai.AsyncOpenAI,
     choice = chat_completion.choices[0]
     assert choice.finish_reason == "length"
     assert chat_completion.usage == openai.types.CompletionUsage(
-        completion_tokens=10, prompt_tokens=6299, total_tokens=6309)
+        completion_tokens=10, prompt_tokens=6287, total_tokens=6297)
 
     message = choice.message
     message = chat_completion.choices[0].message
@@ -180,7 +180,7 @@ async def test_single_chat_session_video_base64encoded(
     choice = chat_completion.choices[0]
     assert choice.finish_reason == "length"
     assert chat_completion.usage == openai.types.CompletionUsage(
-        completion_tokens=10, prompt_tokens=6299, total_tokens=6309)
+        completion_tokens=10, prompt_tokens=6287, total_tokens=6297)
 
     message = choice.message
     message = chat_completion.choices[0].message

tests/entrypoints/test_chat_utils.py

@@ -4,13 +4,16 @@ import warnings
 from typing import Optional
 
 import pytest
+from packaging.version import Version
+from transformers import __version__ as TRANSFORMERS_VERSION
 
 from vllm.assets.image import ImageAsset
 from vllm.config import ModelConfig
 from vllm.entrypoints.chat_utils import (_try_extract_ast, load_chat_template,
                                          parse_chat_messages,
                                          parse_chat_messages_futures,
-                                         resolve_chat_template_content_format)
+                                         resolve_chat_template_content_format,
+                                         resolve_hf_chat_template)
 from vllm.entrypoints.llm import apply_hf_chat_template
 from vllm.multimodal import MultiModalDataDict
 from vllm.multimodal.utils import encode_image_base64
@@ -23,8 +26,10 @@ EXAMPLES_DIR = VLLM_PATH / "examples"
 PHI3V_MODEL_ID = "microsoft/Phi-3.5-vision-instruct"
 ULTRAVOX_MODEL_ID = "fixie-ai/ultravox-v0_5-llama-3_2-1b"
 QWEN2VL_MODEL_ID = "Qwen/Qwen2-VL-2B-Instruct"
+QWEN25VL_MODEL_ID = "Qwen/Qwen2.5-VL-3B-Instruct"
 MLLAMA_MODEL_ID = "meta-llama/Llama-3.2-11B-Vision-Instruct"
 LLAMA_GUARD_MODEL_ID = "meta-llama/Llama-Guard-3-1B"
+HERMES_MODEL_ID = "NousResearch/Hermes-3-Llama-3.1-8B"
 
 
 @pytest.fixture(scope="function")
@@ -703,25 +708,27 @@ def test_multimodal_image_parsing_matches_hf(model, image_url):
 
     vllm_result = apply_hf_chat_template(
         tokenizer,
+        trust_remote_code=model_config.trust_remote_code,
         conversation=conversation,
         chat_template=None,
+        tools=None,
         add_generation_prompt=True,
     )
 
     assert hf_result == vllm_result
 
 
-# yapf: disable
 @pytest.mark.parametrize(
-    ("model", "expected_format"),
-    [(PHI3V_MODEL_ID, "string"),
-     (QWEN2VL_MODEL_ID, "openai"),
-     (ULTRAVOX_MODEL_ID, "string"),
-     (MLLAMA_MODEL_ID, "openai"),
-     (LLAMA_GUARD_MODEL_ID, "openai")],
-)
-# yapf: enable
-def test_resolve_content_format_hf_defined(model, expected_format):
+    "model",
+    [
+        QWEN2VL_MODEL_ID,  # tokenizer.chat_template is of type str
+        HERMES_MODEL_ID,  # tokenizer.chat_template is of type dict
+    ])
+@pytest.mark.parametrize("use_tools", [True, False])
+def test_resolve_hf_chat_template(sample_json_schema, model, use_tools):
+    """checks that chat_template is a dict type for HF models."""
+
+    # Build the tokenizer group and grab the underlying tokenizer
     tokenizer_group = TokenizerGroup(
         model,
         enable_lora=False,
@@ -730,7 +737,56 @@ def test_resolve_content_format_hf_defined(model, expected_format):
     )
     tokenizer = tokenizer_group.tokenizer
 
-    chat_template = tokenizer.chat_template
+    tools = [{
+        "type": "function",
+        "function": {
+            "name": "dummy_function_name",
+            "description": "This is a dummy function",
+            "parameters": sample_json_schema
+        }
+    }] if use_tools else None
+
+    # Test detecting the tokenizer's chat_template
+    chat_template = resolve_hf_chat_template(
+        tokenizer,
+        chat_template=None,
+        tools=tools,
+        trust_remote_code=True,
+    )
+    assert isinstance(chat_template, str)
+
+
+# yapf: disable
+@pytest.mark.parametrize(
+    ("model", "expected_format"),
+    [(PHI3V_MODEL_ID, "string"),
+     (QWEN2VL_MODEL_ID, "openai"),
+     (QWEN25VL_MODEL_ID, "openai"),
+     (ULTRAVOX_MODEL_ID, "string"),
+     (MLLAMA_MODEL_ID, "openai"),
+     (LLAMA_GUARD_MODEL_ID, "openai")],
+)
+# yapf: enable
+def test_resolve_content_format_hf_defined(model, expected_format):
+    if model == QWEN25VL_MODEL_ID and Version(TRANSFORMERS_VERSION) < Version(
+            "4.49.0"):
+        pytest.skip("Qwen2.5-VL requires transformers>=4.49.0")
+
+    tokenizer_group = TokenizerGroup(
+        model,
+        enable_lora=False,
+        max_num_seqs=5,
+        max_input_length=None,
+    )
+    tokenizer = tokenizer_group.tokenizer
+
+    # Test detecting the tokenizer's chat_template
+    chat_template = resolve_hf_chat_template(
+        tokenizer,
+        chat_template=None,
+        tools=None,
+        trust_remote_code=True,
+    )
     assert isinstance(chat_template, str)
 
     print("[TEXT]")
@@ -740,8 +796,10 @@ def test_resolve_content_format_hf_defined(model, expected_format):
 
     resolved_format = resolve_chat_template_content_format(
         None,  # Test detecting the tokenizer's chat_template
+        None,
         "auto",
         tokenizer,
+        trust_remote_code=True,
     )
 
     assert resolved_format == expected_format
@@ -791,8 +849,10 @@ def test_resolve_content_format_examples(template_path, expected_format):
 
     resolved_format = resolve_chat_template_content_format(
         chat_template,
+        None,
         "auto",
         dummy_tokenizer,
+        trust_remote_code=True,
     )
 
     assert resolved_format == expected_format

tests/fastsafetensors_loader/test_fastsafetensors_loader.py

@@ -0,0 +1,22 @@
+# SPDX-License-Identifier: Apache-2.0
+
+from vllm import SamplingParams
+from vllm.config import LoadFormat
+
+test_model = "openai-community/gpt2"
+
+prompts = [
+    "Hello, my name is",
+    "The president of the United States is",
+    "The capital of France is",
+    "The future of AI is",
+]
+# Create a sampling params object.
+sampling_params = SamplingParams(temperature=0.8, top_p=0.95, seed=0)
+
+
+def test_model_loader_download_files(vllm_runner):
+    with vllm_runner(test_model,
+                     load_format=LoadFormat.FASTSAFETENSORS) as llm:
+        deserialized_outputs = llm.generate(prompts, sampling_params)
+        assert deserialized_outputs

tests/fastsafetensors_loader/test_weight_utils.py

@@ -0,0 +1,46 @@
+# SPDX-License-Identifier: Apache-2.0
+
+import glob
+import tempfile
+
+import huggingface_hub.constants
+import torch
+
+from vllm.model_executor.model_loader.weight_utils import (
+    download_weights_from_hf, fastsafetensors_weights_iterator,
+    safetensors_weights_iterator)
+
+
+def test_fastsafetensors_model_loader():
+    with tempfile.TemporaryDirectory() as tmpdir:
+        huggingface_hub.constants.HF_HUB_OFFLINE = False
+        download_weights_from_hf("openai-community/gpt2",
+                                 allow_patterns=["*.safetensors"],
+                                 cache_dir=tmpdir)
+        safetensors = glob.glob(f"{tmpdir}/**/*.safetensors", recursive=True)
+        assert len(safetensors) > 0
+
+        fastsafetensors_tensors = {}
+        hf_safetensors_tensors = {}
+
+        for name, tensor in fastsafetensors_weights_iterator(
+                safetensors, True):
+            fastsafetensors_tensors[name] = tensor
+
+        for name, tensor in safetensors_weights_iterator(safetensors, True):
+            hf_safetensors_tensors[name] = tensor
+
+        assert len(fastsafetensors_tensors) == len(hf_safetensors_tensors)
+
+        for name, fastsafetensors_tensor in fastsafetensors_tensors.items():
+            fastsafetensors_tensor = fastsafetensors_tensor.to('cpu')
+            assert fastsafetensors_tensor.dtype == hf_safetensors_tensors[
+                name].dtype
+            assert fastsafetensors_tensor.shape == hf_safetensors_tensors[
+                name].shape
+            assert torch.all(
+                fastsafetensors_tensor.eq(hf_safetensors_tensors[name]))
+
+
+if __name__ == "__main__":
+    test_fastsafetensors_model_loader()

tests/kernels/test_cache.py

@@ -749,3 +749,72 @@ def test_gather_cache_mla(kv_lora_rank, qk_rope_head_dim, block_size,
 
     ops.gather_cache(src_cache, dst, block_table, cu_seq_lens, batch_size)
     torch.testing.assert_close(dst, expected)
+
+
+@pytest.mark.parametrize("kv_lora_rank", KV_LORA_RANKS)
+@pytest.mark.parametrize("qk_rope_head_dim", QK_ROPE_HEAD_DIMS)
+@pytest.mark.parametrize("num_tokens", NUM_TOKENS_MLA)
+@pytest.mark.parametrize("block_size", BLOCK_SIZES_MLA)
+@pytest.mark.parametrize("num_blocks", NUM_BLOCKS_MLA)
+@pytest.mark.parametrize("dtype", DTYPES)
+@pytest.mark.parametrize("seed", SEEDS)
+@pytest.mark.cpu_model
+@pytest.mark.skipif(not current_platform.is_cpu(), reason="CPU only")
+@torch.inference_mode()
+def test_concat_and_cache_mla_cpu(
+    kv_lora_rank: int,
+    qk_rope_head_dim: int,
+    num_tokens: int,
+    block_size: int,
+    num_blocks: int,
+    dtype: torch.dtype,
+    seed: int,
+) -> None:
+    device = "cpu"
+    kv_cache_dtype = "auto"
+    current_platform.seed_everything(seed)
+    torch.set_default_device(device)
+
+    total_slots = num_blocks * block_size
+    slot_mapping_lst = random.sample(range(total_slots), num_tokens)
+    slot_mapping = torch.tensor(slot_mapping_lst,
+                                dtype=torch.long,
+                                device=device)
+
+    kv_c = torch.randn(num_tokens, kv_lora_rank, dtype=dtype, device=device)
+    k_pe = torch.randn(num_tokens,
+                       qk_rope_head_dim,
+                       dtype=dtype,
+                       device=device)
+    entry_size = kv_lora_rank + qk_rope_head_dim
+
+    scale = torch.tensor(0.1, dtype=torch.float32, device=device)
+    kv_cache = _create_mla_cache(num_blocks, block_size, entry_size, dtype,
+                                 kv_cache_dtype, device)
+    ref_temp = torch.zeros(*kv_cache.shape, dtype=dtype, device=device)
+
+    for i in range(num_tokens):
+        slot = slot_mapping[i].item()
+        block_idx = slot // block_size
+        block_offset = slot % block_size
+        ref_temp[block_idx, block_offset, :kv_lora_rank] = kv_c[i]
+        ref_temp[block_idx, block_offset, kv_lora_rank:] = k_pe[i]
+
+    if kv_cache_dtype == "fp8":
+        ref_kv_cache = torch.empty_like(ref_temp, dtype=kv_cache.dtype)
+        ops.convert_fp8(ref_kv_cache,
+                        ref_temp,
+                        scale.item(),
+                        kv_dtype=kv_cache_dtype)
+    else:
+        ref_kv_cache = ref_temp
+
+    opcheck(
+        torch.ops._C_cache_ops.concat_and_cache_mla,
+        (kv_c, k_pe, kv_cache, slot_mapping, kv_cache_dtype, scale),
+        test_utils=DEFAULT_OPCHECK_TEST_UTILS,
+    )
+
+    ops.concat_and_cache_mla(kv_c, k_pe, kv_cache, slot_mapping,
+                             kv_cache_dtype, scale)
+    torch.testing.assert_close(kv_cache, ref_kv_cache)

tests/kernels/test_cutlass.py

@@ -3,6 +3,7 @@
 
 Run `pytest tests/kernels/test_cutlass.py`.
 """
+import random
 
 import pytest
 import torch
@@ -507,3 +508,136 @@ def test_cutlass_cuda_graph(per_act_token: bool, per_out_ch: bool):
 
 def test_cutlass_support_opcheck():
     opcheck(torch.ops._C.cutlass_scaled_mm_supports_fp8, (capability, ))
+
+
+@pytest.mark.parametrize("num_experts", [8, 64])
+@pytest.mark.parametrize("per_act_token", [True, False])
+@pytest.mark.parametrize("per_out_ch", [True, False])
+@pytest.mark.parametrize("use_bias", [False])
+@pytest.mark.skipif(
+    (lambda x: x is None or not ops.cutlass_group_gemm_supported(x.to_int()))(
+        current_platform.get_device_capability()),
+    reason="Grouped gemm is not supported on this GPU type.")
+def test_cutlass_fp8_group_gemm(num_experts: int, per_act_token: bool,
+                                per_out_ch: bool, use_bias: bool):
+
+    # Device and dtype setup
+    device = "cuda"
+    out_dtype = torch.half
+
+    # Create separate A, B, C tensors for each group
+    a_tensors = []
+    b_tensors = []
+    a_scales_tensors = []
+    b_scales_tensors = []
+    baseline_tensors = []
+
+    expert_offsets = torch.zeros((num_experts + 1),
+                                 device=device,
+                                 dtype=torch.int32)
+
+    problem_sizes = torch.zeros((num_experts, 3),
+                                device=device,
+                                dtype=torch.int32)
+
+    if not per_act_token:
+        one_scale_a = torch.randn((1, 1), device=device, dtype=torch.float32)
+
+    alignment = 16  # 128 // 8
+    # For variation, each group has dimensions
+    n_g = alignment * random.randint(1, 64)
+    k_g = alignment * random.randint(1, 64)
+    for g in range(num_experts):
+        m_g = alignment * random.randint(1, 64)
+
+        expert_offsets[g + 1] = expert_offsets[g] + m_g
+        problem_sizes[g][0] = m_g
+        problem_sizes[g][1] = n_g
+        problem_sizes[g][2] = k_g
+
+        m_a_scales = m_g if per_act_token else 1
+        n_b_scales = n_g if per_out_ch else 1
+
+        print("shape:", m_g, n_g, k_g)
+
+        # Create group-specific A and B (FP8) and output (FP16/FP32)
+        a_g = to_fp8(torch.randn((m_g, k_g), device=device))
+        b_g = to_fp8(torch.randn((n_g, k_g), device=device).t())
+        a_tensors.append(a_g)
+        b_tensors.append(b_g)
+
+        # Set up A/B scales
+        scale_b = torch.randn((1, n_b_scales),
+                              device=device,
+                              dtype=torch.float32)
+        b_scales_tensors.append(scale_b)
+
+        if per_act_token:
+            scale_a = torch.randn((m_a_scales, 1),
+                                  device=device,
+                                  dtype=torch.float32)
+            a_scales_tensors.append(scale_a)
+        else:
+            scale_a = one_scale_a
+
+        # Compute baseline result for this group
+        baseline_g = baseline_scaled_mm(a_g, b_g, scale_a, scale_b, out_dtype,
+                                        None)
+        baseline_tensors.append(baseline_g)
+
+    a_tensors_stacked = torch.empty((expert_offsets[num_experts], k_g),
+                                    device=device,
+                                    dtype=torch.float8_e4m3fn)
+    b_tensors_stacked = torch.empty((num_experts, n_g, k_g),
+                                    device=device,
+                                    dtype=torch.float8_e4m3fn)
+
+    for g in range(num_experts):
+        a_tensors_stacked[expert_offsets[g]:expert_offsets[g +
+                                                           1]] = a_tensors[g]
+        b_tensors_stacked[g] = b_tensors[g].t()
+    b_tensors_stacked = b_tensors_stacked.transpose(1, 2)
+
+    if per_act_token:
+        a_scales_tensors_stacked = torch.empty(
+            (expert_offsets[num_experts], 1),
+            device=device,
+            dtype=torch.float32)
+        for g in range(num_experts):
+            a_scales_tensors_stacked[
+                expert_offsets[g]:expert_offsets[g + 1]] = a_scales_tensors[g]
+    else:
+        a_scales_tensors_stacked = one_scale_a
+
+    b_scales_tensors_stacked = torch.empty((num_experts, n_b_scales),
+                                           device=device,
+                                           dtype=torch.float32)
+    for g in range(num_experts):
+        b_scales_tensors_stacked[g] = b_scales_tensors[g]
+
+    out_tensors_stacked = torch.zeros((expert_offsets[num_experts], n_g),
+                                      device=device,
+                                      dtype=out_dtype)
+
+    ab_strides = torch.full((num_experts, ),
+                            a_tensors_stacked.stride(0),
+                            device="cuda",
+                            dtype=torch.int64)
+    c_strides = torch.full((num_experts, ),
+                           out_tensors_stacked.stride(0),
+                           device="cuda",
+                           dtype=torch.int64)
+
+    ops.cutlass_moe_mm(out_tensors_stacked, a_tensors_stacked,
+                       b_tensors_stacked, a_scales_tensors_stacked,
+                       b_scales_tensors_stacked, expert_offsets[:-1],
+                       problem_sizes, ab_strides, ab_strides, c_strides)
+
+    # Validate each group's result against the baseline
+    for g in range(num_experts):
+        baseline = baseline_tensors[g]
+        c = out_tensors_stacked[expert_offsets[g]:expert_offsets[g + 1]]
+        print(baseline)
+        print(c)
+        print("*")
+        torch.testing.assert_close(c, baseline, rtol=1e-2, atol=5e-4)

tests/kernels/test_cutlass_moe.py

@@ -0,0 +1,244 @@
+# SPDX-License-Identifier: Apache-2.0
+import pytest
+import torch
+
+from vllm import _custom_ops as ops
+from vllm.config import ParallelConfig, VllmConfig, set_current_vllm_config
+from vllm.model_executor.layers.fused_moe.fused_moe import (cutlass_moe_fp8,
+                                                            fused_experts,
+                                                            fused_topk)
+from vllm.platforms import current_platform
+
+NUM_EXPERTS = [40, 64]
+TOP_KS = [6, 8]
+
+
+def run(a: torch.Tensor, a_scale: torch.Tensor, w1_q: torch.Tensor,
+        w2_q: torch.Tensor, w1_scale: torch.Tensor, w2_scale: torch.Tensor,
+        topk_weights: torch.Tensor, topk_ids: torch.Tensor,
+        ab_strides1: torch.Tensor, c_strides1: torch.Tensor,
+        ab_strides2: torch.Tensor, c_strides2: torch.Tensor):
+    with set_current_vllm_config(
+            VllmConfig(parallel_config=ParallelConfig(
+                pipeline_parallel_size=1))):
+        return cutlass_moe_fp8(a,
+                               w1_q,
+                               w2_q,
+                               w1_scale,
+                               w2_scale,
+                               topk_weights,
+                               topk_ids,
+                               ab_strides1,
+                               c_strides1,
+                               ab_strides2,
+                               c_strides2,
+                               a1_scale=a_scale)
+
+
+@pytest.mark.parametrize("m", [2, 64, 224])
+@pytest.mark.parametrize("n", [1024, 3072])
+@pytest.mark.parametrize("k", [1024, 1536])
+@pytest.mark.parametrize("e", NUM_EXPERTS)
+@pytest.mark.parametrize("topk", TOP_KS)
+@pytest.mark.parametrize("per_act_token", [True, False])
+@pytest.mark.parametrize("per_out_ch", [True, False])
+@pytest.mark.skipif(
+    (lambda x: x is None or not ops.cutlass_group_gemm_supported(x.to_int()))(
+        current_platform.get_device_capability()),
+    reason="Grouped gemm is not supported on this GPU type.")
+def test_cutlass_moe_no_graph(
+    m: int,
+    n: int,
+    k: int,
+    e: int,
+    topk: int,
+    per_act_token: bool,
+    per_out_ch: bool,
+):
+    current_platform.seed_everything(7)
+    with set_current_vllm_config(
+            VllmConfig(parallel_config=ParallelConfig(
+                pipeline_parallel_size=1))):
+
+        dtype = torch.half
+
+        a = torch.randn((m, k), device="cuda", dtype=dtype) / 10
+        w1 = torch.randn((e, 2 * n, k), device="cuda", dtype=dtype) / 10
+        w2 = torch.randn((e, k, n), device="cuda", dtype=dtype) / 10
+
+        # Get the right scale for tests.
+        _, a_scale1 = ops.scaled_fp8_quant(
+            a, use_per_token_if_dynamic=per_act_token)
+        a_q, _ = ops.scaled_fp8_quant(a,
+                                      a_scale1,
+                                      use_per_token_if_dynamic=per_act_token)
+
+        a_d = a_q.float().mul(a_scale1).to(dtype)
+
+        n_b_scales = 2 * n if per_out_ch else 1
+        k_b_scales = k if per_out_ch else 1
+
+        w1_q = torch.empty((e, 2 * n, k),
+                           device="cuda",
+                           dtype=torch.float8_e4m3fn)
+        w2_q = torch.empty((e, k, n), device="cuda", dtype=torch.float8_e4m3fn)
+        w1_scale = torch.empty((e, n_b_scales, 1),
+                               device="cuda",
+                               dtype=torch.float32)
+        w2_scale = torch.empty((e, k_b_scales, 1),
+                               device="cuda",
+                               dtype=torch.float32)
+
+        ab_strides1 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+        c_strides1 = torch.full((e, ), 2 * n, device="cuda", dtype=torch.int64)
+        ab_strides2 = torch.full((e, ), n, device="cuda", dtype=torch.int64)
+        c_strides2 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+
+        for expert in range(e):
+            w1_q[expert], w1_scale[expert] = ops.scaled_fp8_quant(
+                w1[expert], use_per_token_if_dynamic=per_out_ch)
+            w2_q[expert], w2_scale[expert] = ops.scaled_fp8_quant(
+                w2[expert], use_per_token_if_dynamic=per_out_ch)
+        w1_q = w1_q.transpose(1, 2)
+        w2_q = w2_q.transpose(1, 2)
+
+        ab_strides1 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+        c_strides1 = torch.full((e, ), 2 * n, device="cuda", dtype=torch.int64)
+        ab_strides2 = torch.full((e, ), n, device="cuda", dtype=torch.int64)
+        c_strides2 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+
+        w1_d = torch.empty_like(w1)
+        w2_d = torch.empty_like(w2)
+        for expert in range(e):
+            w1_d[expert] = (w1_q[expert].t().float() * w1_scale[expert]).half()
+            w2_d[expert] = (w2_q[expert].t().float() * w2_scale[expert]).half()
+
+        score = torch.randn((m, e), device="cuda", dtype=dtype)
+        topk_weights, topk_ids = fused_topk(a, score, topk, renormalize=False)
+
+        triton_output = fused_experts(a_d, w1_d, w2_d, topk_weights, topk_ids)
+
+        cutlass_output = cutlass_moe_fp8(a,
+                                         w1_q,
+                                         w2_q,
+                                         w1_scale,
+                                         w2_scale,
+                                         topk_weights,
+                                         topk_ids,
+                                         ab_strides1,
+                                         c_strides1,
+                                         ab_strides2,
+                                         c_strides2,
+                                         a1_scale=a_scale1)
+
+        print(triton_output)
+        print(cutlass_output)
+        print("*")
+
+        torch.testing.assert_close(triton_output,
+                                   cutlass_output,
+                                   atol=5e-2,
+                                   rtol=1e-2)
+
+
+@pytest.mark.parametrize("m", [2, 64, 224])
+@pytest.mark.parametrize("n", [1024, 3072])
+@pytest.mark.parametrize("k", [1024, 1536])
+@pytest.mark.parametrize("e", NUM_EXPERTS)
+@pytest.mark.parametrize("topk", TOP_KS)
+@pytest.mark.parametrize("per_act_token", [True, False])
+@pytest.mark.parametrize("per_out_ch", [True, False])
+@pytest.mark.skipif(
+    (lambda x: x is None or not ops.cutlass_group_gemm_supported(x.to_int()))(
+        current_platform.get_device_capability()),
+    reason="Grouped gemm is not supported on this GPU type.")
+def test_cutlass_moe_cuda_graph(
+    m: int,
+    n: int,
+    k: int,
+    e: int,
+    topk: int,
+    per_act_token: bool,
+    per_out_ch: bool,
+):
+    current_platform.seed_everything(7)
+    with set_current_vllm_config(
+            VllmConfig(parallel_config=ParallelConfig(
+                pipeline_parallel_size=1))):
+
+        dtype = torch.half
+
+        a = torch.randn((m, k), device="cuda", dtype=dtype) / 10
+        w1 = torch.randn((e, 2 * n, k), device="cuda", dtype=dtype) / 10
+        w2 = torch.randn((e, k, n), device="cuda", dtype=dtype) / 10
+
+        # Get the right scale for tests.
+        _, a_scale1 = ops.scaled_fp8_quant(
+            a, use_per_token_if_dynamic=per_act_token)
+        a_q, _ = ops.scaled_fp8_quant(a,
+                                      a_scale1,
+                                      use_per_token_if_dynamic=per_act_token)
+
+        a_d = a_q.float().mul(a_scale1).to(dtype)
+
+        n_b_scales = 2 * n if per_out_ch else 1
+        k_b_scales = k if per_out_ch else 1
+
+        w1_q = torch.empty((e, 2 * n, k),
+                           device="cuda",
+                           dtype=torch.float8_e4m3fn)
+        w2_q = torch.empty((e, k, n), device="cuda", dtype=torch.float8_e4m3fn)
+        w1_scale = torch.empty((e, n_b_scales, 1),
+                               device="cuda",
+                               dtype=torch.float32)
+        w2_scale = torch.empty((e, k_b_scales, 1),
+                               device="cuda",
+                               dtype=torch.float32)
+
+        ab_strides1 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+        c_strides1 = torch.full((e, ), 2 * n, device="cuda", dtype=torch.int64)
+        ab_strides2 = torch.full((e, ), n, device="cuda", dtype=torch.int64)
+        c_strides2 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+
+        for expert in range(e):
+            w1_q[expert], w1_scale[expert] = ops.scaled_fp8_quant(
+                w1[expert], use_per_token_if_dynamic=per_out_ch)
+            w2_q[expert], w2_scale[expert] = ops.scaled_fp8_quant(
+                w2[expert], use_per_token_if_dynamic=per_out_ch)
+        w1_q = w1_q.transpose(1, 2)
+        w2_q = w2_q.transpose(1, 2)
+
+        ab_strides1 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+        c_strides1 = torch.full((e, ), 2 * n, device="cuda", dtype=torch.int64)
+        ab_strides2 = torch.full((e, ), n, device="cuda", dtype=torch.int64)
+        c_strides2 = torch.full((e, ), k, device="cuda", dtype=torch.int64)
+
+        w1_d = torch.empty_like(w1)
+        w2_d = torch.empty_like(w2)
+        for expert in range(e):
+            w1_d[expert] = (w1_q[expert].t().float() * w1_scale[expert]).half()
+            w2_d[expert] = (w2_q[expert].t().float() * w2_scale[expert]).half()
+
+        score = torch.randn((m, e), device="cuda", dtype=dtype)
+        topk_weights, topk_ids = fused_topk(a, score, topk, renormalize=False)
+
+        triton_output = fused_experts(a_d, w1_d, w2_d, topk_weights, topk_ids)
+
+        stream = torch.cuda.Stream()
+        graph = torch.cuda.CUDAGraph()
+        with torch.cuda.graph(graph, stream=stream):
+            cutlass_output = run(a, a_scale1, w1_q, w2_q, w1_scale, w2_scale,
+                                 topk_weights, topk_ids, ab_strides1,
+                                 c_strides1, ab_strides2, c_strides2)
+        torch.cuda.synchronize()
+        graph.replay()
+        torch.cuda.synchronize()
+
+        print(triton_output)
+        print(cutlass_output)
+        print("*")
+
+        torch.testing.assert_close(triton_output,
+                                   cutlass_output,
+                                   atol=9e-2,
+                                   rtol=1e-2)

tests/kernels/test_marlin_gemm.py

@@ -606,6 +606,51 @@ def test_marlin_qqq_gemm(
     assert max_diff < 0.04
 
 
+def test_marlin_gemm_subset_input():
+    quant_type = scalar_types.uint4b8
+    group_size = 128
+
+    size_m, size_k, size_n = 32, 1024, 2048
+    big_m = size_m * 2
+    big_k = size_k * 2
+
+    a_input = rand_data((big_m, big_k))[8:size_m + 8, 8:size_k + 8]
+    b_weight = rand_data((size_k, size_n))
+
+    w_ref, marlin_q_w, marlin_s, g_idx, sort_indices, _ = marlin_quantize(
+        b_weight, quant_type, group_size, False)
+
+    marlin_zp = marlin_make_empty_g_idx(marlin_s.device)
+    workspace = MarlinWorkspace(size_n, GPTQ_MARLIN_MIN_THREAD_N,
+                                GPTQ_MARLIN_MAX_PARALLEL)
+
+    output = ops.gptq_marlin_gemm(
+        a_input,
+        marlin_q_w,
+        marlin_s,
+        marlin_zp,
+        g_idx,
+        sort_indices,
+        workspace.scratch,
+        quant_type,
+        a_input.shape[0],
+        b_weight.shape[1],
+        a_input.shape[1],
+        is_k_full=True,
+        has_zp=False,
+        use_atomic_add=False,
+        use_fp32_reduce=True,
+        is_zp_float=False,
+    )
+    output_ref = torch.matmul(a_input, w_ref)
+
+    torch.cuda.synchronize()
+
+    max_diff = compute_max_diff(output, output_ref)
+
+    assert max_diff < 0.04
+
+
 def test_marlin_gemm_opcheck():
     size_m = 2048
     size_n = 4096

tests/kernels/test_mla_decode_cpu.py

@@ -0,0 +1,94 @@
+# SPDX-License-Identifier: Apache-2.0
+import pytest
+import torch
+import torch.nn.functional as F
+from torch import Tensor
+
+import vllm._custom_ops as ops
+from vllm.platforms import current_platform
+
+
+def cdiv(a, b):
+    return (a + b - 1) // b
+
+
+def ref_mla(
+        out: Tensor,  # (bs, num_heads, v_head_dim)
+        query: Tensor,  # (bs, num_heads, head_dim)
+        kv_cache: Tensor,  # (num_blocks, block_size, head_dim)
+        scale: float,
+        block_tables: Tensor,  # (bs, max_num_blocks)
+        seq_lens: Tensor,  # (bs,)
+):
+    bs, num_heads, v_head_dim = out.shape
+    head_dim = query.shape[2]
+
+    for i in range(bs):
+        # gather and flatten KV-cache
+        kv = kv_cache[
+            block_tables[i]]  # (max_num_blocks, block_size, head_dim)
+        kv = kv.view(1, -1,
+                     head_dim)[:, :seq_lens[i]]  # (1, seq_len, head_dim)
+        v = kv[:, :, :v_head_dim]
+
+        q = query[i].view(num_heads, 1, head_dim)
+        o = F.scaled_dot_product_attention(q,
+                                           kv,
+                                           v,
+                                           scale=scale,
+                                           enable_gqa=True)
+        out[i] = o.view(num_heads, v_head_dim)
+
+    return out
+
+
+@pytest.mark.parametrize("bs", [4])
+@pytest.mark.parametrize("mean_seq_len", [256])
+@pytest.mark.parametrize("h_q", [16])
+@pytest.mark.parametrize("d", [576])
+@pytest.mark.parametrize("dv", [512])
+@pytest.mark.parametrize("block_size", [16])
+@pytest.mark.parametrize("dtype", [torch.float, torch.half, torch.bfloat16])
+@pytest.mark.parametrize("varlen", [False, True])
+@pytest.mark.cpu_model
+@pytest.mark.skipif(not current_platform.is_cpu(), reason="CPU only")
+def test_mla_decode_cpu(
+    bs: int,
+    mean_seq_len: int,
+    h_q: int,
+    d: int,
+    dv: int,
+    block_size: int,
+    dtype: torch.dtype,
+    varlen: bool,
+):
+    torch.set_default_dtype(dtype)
+    torch.manual_seed(0)
+
+    scale = d**(-0.5)
+    if varlen:
+        seq_lens = torch.empty(bs).normal_(mean_seq_len, mean_seq_len / 2)
+        seq_lens = seq_lens.clip(2).to(torch.int32)
+    else:
+        seq_lens = torch.full((bs, ), mean_seq_len, dtype=torch.int32)
+    max_seq_len = seq_lens.max().item()
+    seqlen_pad = cdiv(max_seq_len, 256) * 256  # is this necessary?
+
+    q = torch.randn(bs, h_q, d)
+    block_table = torch.arange(bs * seqlen_pad // block_size,
+                               dtype=torch.int32)
+    block_table = block_table.view(bs, seqlen_pad // block_size)
+
+    kv_cache = torch.randn(block_table.numel(), block_size, d)
+    for i, seq_len in enumerate(seq_lens.tolist()):
+        kv_cache.view(bs, seqlen_pad, d)[i, seq_len:] = float("nan")
+
+    out_mla = q.new_zeros(bs, h_q, dv)
+    ops.mla_decode_kvcache_cpu(out_mla, q, kv_cache, scale, block_table,
+                               seq_lens)
+
+    out_ref = q.new_zeros(bs, h_q, dv)
+    ref_mla(out_ref, q, kv_cache, scale, block_table, seq_lens)
+
+    assert not out_mla.isnan().any(), "Likely read out of bounds"
+    torch.testing.assert_close(out_mla, out_ref)

tests/kernels/test_moe.py

@@ -5,6 +5,8 @@ Run `pytest tests/kernels/test_moe.py`.
 """
 import pytest
 import torch
+from torch.nn import Parameter
+from torch.nn import functional as F
 from transformers import MixtralConfig
 from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
 
@@ -37,6 +39,7 @@ TOP_KS = [2, 6]
 @pytest.mark.parametrize("topk", TOP_KS)
 @pytest.mark.parametrize("ep_size", EP_SIZE)
 @pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16])
+@pytest.mark.parametrize("padding", [True, False])
 def test_fused_moe(
     m: int,
     n: int,
@@ -45,6 +48,7 @@ def test_fused_moe(
     topk: int,
     ep_size: int,
     dtype: torch.dtype,
+    padding: bool,
 ):
     a = torch.randn((m, k), device="cuda", dtype=dtype) / 10
     w1 = torch.randn((e, 2 * n, k), device="cuda", dtype=dtype) / 10
@@ -65,16 +69,7 @@ def test_fused_moe(
     else:
         e_map = None
 
-    triton_output = fused_moe(a,
-                              w1,
-                              w2,
-                              score,
-                              topk,
-                              global_num_experts=e,
-                              expert_map=e_map,
-                              renormalize=False)
     torch_output = torch_moe(a, w1, w2, score, topk, e_map)
-    torch.testing.assert_close(triton_output, torch_output, atol=2e-2, rtol=0)
     iterative_output = iterative_moe(a,
                                      w1,
                                      w2,
@@ -83,6 +78,23 @@ def test_fused_moe(
                                      global_num_experts=e,
                                      expert_map=e_map,
                                      renormalize=False)
+
+    # Pad the weight if moe padding is enabled
+    if padding:
+        w1 = F.pad(w1, (0, 128), "constant", 0)[..., 0:-128]
+        torch.cuda.empty_cache()
+        w2 = F.pad(w2, (0, 128), "constant", 0)[..., 0:-128]
+        torch.cuda.empty_cache()
+
+    triton_output = fused_moe(a,
+                              w1,
+                              w2,
+                              score,
+                              topk,
+                              global_num_experts=e,
+                              expert_map=e_map,
+                              renormalize=False)
+    torch.testing.assert_close(triton_output, torch_output, atol=2e-2, rtol=0)
     torch.testing.assert_close(iterative_output,
                                torch_output,
                                atol=2e-2,
@@ -202,11 +214,18 @@ def test_fused_moe_wn16(m: int, n: int, k: int, e: int, topk: int,
 
 @pytest.mark.parametrize("dtype",
                          [torch.float32, torch.float16, torch.bfloat16])
+@pytest.mark.parametrize("padding", [True, False])
+@pytest.mark.parametrize(
+    "use_rocm_aiter", [True, False] if current_platform.is_rocm() else [False])
 @torch.inference_mode()
-def test_mixtral_moe(dtype: torch.dtype):
+def test_mixtral_moe(dtype: torch.dtype, padding: bool, use_rocm_aiter: bool,
+                     monkeypatch):
     """Make sure our Mixtral MoE implementation agrees with the one from
     huggingface."""
 
+    if use_rocm_aiter:
+        monkeypatch.setenv("VLLM_ROCM_USE_AITER", "1")
+
     # Instantiate our and huggingface's MoE blocks
     config = MixtralConfig()
     hf_moe = MixtralSparseMoeBlock(config).to(dtype).to("cuda")
@@ -233,6 +252,17 @@ def test_mixtral_moe(dtype: torch.dtype):
     # vLLM uses 1D query [num_tokens, hidden_dim]
     vllm_inputs = hf_inputs.flatten(0, 1)
 
+    # Pad the weight if moe padding is enabled
+    if padding:
+        vllm_moe.experts.w13_weight = Parameter(F.pad(
+            vllm_moe.experts.w13_weight, (0, 128), "constant", 0)[..., 0:-128],
+                                                requires_grad=False)
+        torch.cuda.empty_cache()
+        vllm_moe.experts.w2_weight = Parameter(F.pad(
+            vllm_moe.experts.w2_weight, (0, 128), "constant", 0)[..., 0:-128],
+                                               requires_grad=False)
+        torch.cuda.empty_cache()
+
     # Run forward passes for both MoE blocks
     hf_states, _ = hf_moe.forward(hf_inputs)
     vllm_states = vllm_moe.forward(vllm_inputs)
@@ -243,10 +273,18 @@ def test_mixtral_moe(dtype: torch.dtype):
         torch.bfloat16: 1e-2,
     }
 
-    torch.testing.assert_close(hf_states.flatten(0, 1),
-                               vllm_states,
-                               rtol=mixtral_moe_tol[dtype],
-                               atol=mixtral_moe_tol[dtype])
+    if use_rocm_aiter:
+        # The values of rtol and atol are set based on the tests in ROCM AITER package. # noqa: E501
+        # https://github.com/ROCm/aiter/blob/dfed377f4be7da96ca2d75ac0761f569676f7240/op_tests/test_moe.py#L174  # noqa: E501
+        torch.testing.assert_close(hf_states.flatten(0, 1),
+                                   vllm_states,
+                                   rtol=0.01,
+                                   atol=100)
+    else:
+        torch.testing.assert_close(hf_states.flatten(0, 1),
+                                   vllm_states,
+                                   rtol=mixtral_moe_tol[dtype],
+                                   atol=mixtral_moe_tol[dtype])
 
 
 @pytest.mark.parametrize("m", [1, 33, 64, 222])

tests/lora/conftest.py

@@ -241,39 +241,6 @@ def long_context_lora_files_16k_1():
     return snapshot_download(repo_id="SangBinCho/long_context_16k_testing_1")
 
 
-@pytest.fixture(scope="session")
-def long_context_lora_files_16k_2():
-    return snapshot_download(repo_id="SangBinCho/long_context_16k_testing_2")
-
-
-@pytest.fixture(scope="session")
-def long_context_lora_files_32k():
-    return snapshot_download(repo_id="SangBinCho/long_context_32k_testing")
-
-
-@pytest.fixture(scope="session")
-def long_context_infos(long_context_lora_files_16k_1,
-                       long_context_lora_files_16k_2,
-                       long_context_lora_files_32k):
-    cleanup_dist_env_and_memory(shutdown_ray=True)
-    infos: dict[int, ContextInfo] = {}
-    for lora_checkpoint_info in LONG_LORA_INFOS:
-        lora_id = lora_checkpoint_info["lora_id"]
-        if lora_id == 1:
-            lora = long_context_lora_files_16k_1
-        elif lora_id == 2:
-            lora = long_context_lora_files_16k_2
-        elif lora_id == 3:
-            lora = long_context_lora_files_32k
-        else:
-            raise AssertionError("Unknown lora id")
-        infos[lora_id] = {
-            "context_length": lora_checkpoint_info["context_length"],
-            "lora": lora,
-        }
-    return infos
-
-
 @pytest.fixture
 def llama_2_7b_engine_extra_embeddings():
     cleanup_dist_env_and_memory(shutdown_ray=True)