[Core] Update PyTorch to 2.14.0, torchvision to 0.29.0, triton to 3.8.0 (test channel)

.buildkite/hardware_tests/cpu.yaml

@@ -123,9 +123,13 @@ steps:
   - tests/models/language/generation/
   - tests/models/language/pooling/
   - tests/v1/e2e/test_cpu_linear_attn_chunked_prefix.py
+  # RELEASE-ONLY: torch 2.14.0 makes this shard ~2.6x slower than 2.13 (pytest phase
+  # 1473-1593s on 2.13 vs 3969s on 2.14), so the old 50m limit killed it mid-run.
+  # 90m is enough: it now completes in ~66m. Revert once
+  # https://github.com/pytorch/pytorch/issues/193951 is resolved.
   commands:
     - |
-      bash .buildkite/scripts/hardware_ci/run-cpu-test.sh 25m "
+      bash .buildkite/scripts/hardware_ci/run-cpu-test.sh 45m "
       pytest -x -v -s tests/models/language/generation tests/models/language/pooling tests/v1/e2e/test_cpu_linear_attn_chunked_prefix.py -m cpu_model --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB"
   parallelism: 3
 
@@ -191,9 +195,14 @@ steps:
   # - vllm/
   - vllm/model_executor/layers/rotary_embedding
   - tests/models/multimodal/generation/
+  # RELEASE-ONLY: torch 2.14.0 makes these shards up to ~6.6x slower than 2.13
+  # (Multi-Modal 2 pytest phase: 711-781s on 2.13 vs 4924s on 2.14). Shard 1 is the
+  # heaviest and still exceeded 90m, so this is raised to 120m. Shards 2-4 finish
+  # well under it. Revert once
+  # https://github.com/pytorch/pytorch/issues/193951 is resolved.
   commands:
     - |
-      bash .buildkite/scripts/hardware_ci/run-cpu-test.sh 45m "
+      bash .buildkite/scripts/hardware_ci/run-cpu-test.sh 120m "
       pytest -x -v -s tests/models/multimodal/generation --ignore=tests/models/multimodal/generation/test_pixtral.py --ignore=tests/models/multimodal/generation/test_qwen2_5_vl.py -m cpu_model --num-shards=$$BUILDKITE_PARALLEL_JOB_COUNT --shard-id=$$BUILDKITE_PARALLEL_JOB"
   parallelism: 4
 

.buildkite/image_build/image_build_cpu.sh

@@ -101,6 +101,14 @@ else
   docker buildx inspect --bootstrap
 
   # build and push
+  # RELEASE-ONLY (torch 2.14.0): --no-cache is required because the RC is
+  # republished to the PyTorch test channel under an unchanging version string.
+  # torch is installed in the `base-common` stage, upstream of the source COPY,
+  # so its layer key depends only on the base image and requirements/cpu.txt --
+  # neither of which changes on a respin or a rebase. Without this, the ECR
+  # registry cache below restores the layer and torch is never re-resolved
+  # (build 84705: 28 CACHED layers, zero torch downloads). Revert once 2.14.0 is
+  # final and published to PyPI.
   docker buildx build --file docker/Dockerfile.cpu \
     --build-arg max_jobs=16 \
     --build-arg buildkite_commit="$BUILDKITE_COMMIT" \
@@ -109,6 +117,7 @@ else
     "${PYTORCH_NIGHTLY_ARGS[@]}" \
     --tag "$IMAGE" \
     --target vllm-test \
+    --no-cache \
     "${CACHE_FROM_ARGS[@]}" \
     --cache-to "type=registry,ref=${CACHE_TO},mode=max" \
     --push \

.buildkite/scripts/check-ray-compatibility.sh

@@ -18,24 +18,28 @@ trap 'rm -rf "$WORK_DIR"' EXIT
 
 # ── Detect PyTorch index URL ─────────────────────────────────────────────
 
+# Offer the resolver every channel that could serve the installed torch and let
+# uv (run below with --index-strategy unsafe-best-match) pick whichever one
+# actually has the pinned version. For CUDA we pass both the stable and the test
+# channels: release candidates live on the test channel until they are promoted
+# to stable, so a stable-only index breaks RC validation (e.g. a torch==X.Y.Z pin
+# from nixl-cuNN that is only published to /whl/test). Both channels are added
+# unconditionally rather than probing the directory listing, since those listings
+# are not reliably reachable from CI egress.
+TORCH_INDEX_URLS=()
 if python3 -c "import torch; assert torch.version.hip" 2>/dev/null; then
     ROCM_VER=$(python3 -c "import torch; print(torch.version.hip.rsplit('.', 1)[0])")
-    CANDIDATE_URL="https://download.pytorch.org/whl/rocm${ROCM_VER}"
-    if curl -fsSL --head "${CANDIDATE_URL}/" >/dev/null 2>&1; then
-        TORCH_INDEX_URL="${CANDIDATE_URL}"
-    else
-        echo ">>> WARNING: ROCm ${ROCM_VER} wheel index not found at ${CANDIDATE_URL}"
-        echo ">>>          Falling back to default PyPI (resolution may be incomplete)"
-        TORCH_INDEX_URL=""
-    fi
+    TORCH_INDEX_URLS+=("https://download.pytorch.org/whl/rocm${ROCM_VER}")
+    TORCH_INDEX_URLS+=("https://download.pytorch.org/whl/test/rocm${ROCM_VER}")
 else
     if [ "${TORCH_NIGHTLY:-0}" = "1" ]; then
-        TORCH_INDEX_URL="https://download.pytorch.org/whl/nightly/cu130"
+        TORCH_INDEX_URLS+=("https://download.pytorch.org/whl/nightly/cu130")
     else
-        TORCH_INDEX_URL="https://download.pytorch.org/whl/cu130"
+        TORCH_INDEX_URLS+=("https://download.pytorch.org/whl/cu130")
+        TORCH_INDEX_URLS+=("https://download.pytorch.org/whl/test/cu130")
     fi
 fi
-echo ">>> Using PyTorch index: ${TORCH_INDEX_URL:-PyPI default}"
+echo ">>> Using PyTorch indexes: ${TORCH_INDEX_URLS[*]:-PyPI default}"
 
 # Fetch all Ray requirement files used in the LLM depset pipeline
 echo ">>> Fetching Ray requirement files"
@@ -138,9 +142,9 @@ echo ">>> Resolving: Can Ray generate compatible lock files?"
 echo "============================================================"
 
 EXTRA_INDEX_ARGS=()
-if [[ -n "${TORCH_INDEX_URL}" ]]; then
-    EXTRA_INDEX_ARGS+=(--extra-index-url "${TORCH_INDEX_URL}")
-fi
+for INDEX_URL in "${TORCH_INDEX_URLS[@]}"; do
+    EXTRA_INDEX_ARGS+=(--extra-index-url "${INDEX_URL}")
+done
 
 set +e
 uv pip compile \

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

@@ -65,7 +65,11 @@ BUILD_RETRY_WAITS=(10 20 40)  # seconds to wait before retry 1/2/3
 build_log="$(mktemp)"
 attempt=1
 while true; do
-    if docker build --progress plain --tag "$IMAGE_NAME" --target vllm-test \
+    # RELEASE-ONLY (torch 2.14.0): torch installs in the `base-common` stage, whose
+    # layer key does not change when the RC is respun under the same version
+    # string, so the local BuildKit cache would restore a pre-revert wheel.
+    # Revert once 2.14.0 is final and published to PyPI.
+    if docker build --progress plain --no-cache --tag "$IMAGE_NAME" --target vllm-test \
             --build-arg USE_SCCACHE=1 --build-arg SCCACHE_LOCAL_ONLY=1 --build-arg max_jobs=16 \
             -f docker/Dockerfile.cpu . 2>&1 | tee "$build_log"; then
         break

.buildkite/test-amd.yaml

@@ -2531,7 +2531,7 @@ steps:
   # https://github.com/pytorch/ao/issues/2919, we'll have to skip new torchao tests for now
   # we can only upgrade after this is resolved
   # TODO(jerryzh168): resolve the above comment
-  - uv pip install --system torchao==0.17.0
+  - uv pip install --system torchao==0.18.0
   - uv pip install --system conch-triton-kernels
   - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
   - pytest -v -s rocm/test_moe_weight_replay.py
@@ -3886,7 +3886,7 @@ steps:
   - vllm/_aiter_ops.py
   - vllm/platforms/rocm.py
   commands:
-  - uv pip install --system torchao==0.17.0
+  - uv pip install --system torchao==0.18.0
   - uv pip install --system conch-triton-kernels
   - VLLM_TEST_FORCE_LOAD_FORMAT=auto pytest -v -s quantization/ --ignore quantization/test_blackwell_moe.py
 

.buildkite/test_areas/quantization.yaml

@@ -15,7 +15,7 @@ steps:
   - tests/quantization
   commands:
   # Pin torchao to the version compatible with the CI PyTorch and CUDA stack.
-  - uv pip install --system torchao==0.17.0 --index-url https://download.pytorch.org/whl/cu130
+  - uv pip install --system torchao==0.18.0 --index-url https://download.pytorch.org/whl/test/cu130
   - uv pip install --system conch-triton-kernels
   # The SM90-only checkpoint currently contains a removed weight_chan_scale
   # parameter. It was not exercised by the previous L4 job.

CMakeLists.txt

@@ -68,8 +68,8 @@ endif()
 # requirements.txt files and should be kept consistent.  The ROCm torch
 # versions are derived from docker/Dockerfile.rocm
 #
-set(TORCH_SUPPORTED_VERSION_CUDA "2.13.0")
-set(TORCH_SUPPORTED_VERSION_ROCM "2.13.0")
+set(TORCH_SUPPORTED_VERSION_CUDA "2.14.0")
+set(TORCH_SUPPORTED_VERSION_ROCM "2.14.0")
 # TORCH_NIGHTLY=1 builds run against unpinned nightly wheels, so the supported-
 # version check would always warn. Only treat it as a nightly build when the
 # value is exactly "1" (the bootstrap exports TORCH_NIGHTLY=0 by default, which

docker/Dockerfile

@@ -77,7 +77,7 @@ ARG UV_INDEX_URL=${PIP_INDEX_URL}
 ARG UV_EXTRA_INDEX_URL=${PIP_EXTRA_INDEX_URL}
 
 # PyTorch provides its own indexes for standard and nightly builds
-ARG PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl
+ARG PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl/test
 
 # PIP supports multiple authentication schemes, including keyring
 # By parameterizing the PIP_KEYRING_PROVIDER variable and setting it to

docker/Dockerfile.cpu

@@ -29,6 +29,13 @@
 # builder. sccache's cache lives in S3 instead, so it does.
 FROM ubuntu:25.04 AS base
 
+# RELEASE-ONLY (torch 2.14.0): the RC is republished to the PyTorch test channel
+# under an unchanging version string and filename, so uv's persistent cache mount
+# at /root/.cache/uv will happily serve a stale wheel even when the layer is
+# rebuilt with --no-cache. Disable the uv cache so every build re-downloads torch
+# and triton. Remove once 2.14.0 is final and published to PyPI.
+ENV UV_NO_CACHE=1
+
 RUN --mount=type=cache,target=/var/cache/apt,sharing=locked \
     --mount=type=cache,target=/var/lib/apt,sharing=locked \
     apt-get update -y \
@@ -184,7 +191,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
         && uv pip install -r requirements/cpu.txt \
             --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/nightly/cpu; \
     else \
-        uv pip install -r requirements/cpu.txt --torch-backend cpu; \
+        uv pip install -r requirements/cpu.txt \
+            --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/test/cpu; \
     fi
 
 ######################### x86_64 BASE IMAGE #########################
@@ -315,7 +323,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
         && uv pip install -r requirements/build/cpu.txt \
             --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/nightly/cpu; \
     else \
-        uv pip install -r requirements/build/cpu.txt --torch-backend cpu; \
+        uv pip install -r requirements/build/cpu.txt \
+            --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/test/cpu; \
     fi
 
 COPY . .
@@ -375,7 +384,7 @@ WORKDIR /vllm-workspace
 
 # Test requirements are compiled from requirements/test/cuda.in into
 # requirements/test/cpu.txt by the pip-compile-cpu pre-commit hook, which
-# resolves CPU wheels via uv's --torch-backend cpu.
+# resolves CPU wheels from the torch test index.
 COPY requirements/test/cpu.txt requirements/test/cpu.txt
 
 # cpu.txt is compiled for x86_64, so platform markers are resolved away. Drop
@@ -393,7 +402,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
         && uv pip install -r requirements/test/cpu.txt \
             --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/nightly/cpu; \
     else \
-        uv pip install -r requirements/test/cpu.txt --torch-backend cpu; \
+        uv pip install -r requirements/test/cpu.txt \
+            --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/test/cpu; \
     fi
 
 ######################### DEV IMAGE #########################
@@ -430,7 +440,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
         uv pip install -r requirements/test/cpu.txt \
             --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/nightly/cpu; \
     else \
-        uv pip install -r requirements/test/cpu.txt --torch-backend cpu; \
+        uv pip install -r requirements/test/cpu.txt \
+            --extra-index-url ${PYTORCH_CPU_INDEX_BASE_URL}/test/cpu; \
     fi && \
     pre-commit install --hook-type pre-commit --hook-type commit-msg
 

docker/Dockerfile.s390x

@@ -70,13 +70,13 @@ RUN curl -LO https://github.com/gperftools/gperftools/releases/download/gperftoo
 
 FROM python-install AS torch-vision
 # Install torchvision
-ARG TORCH_VISION_VERSION=v0.28.0
+ARG TORCH_VISION_VERSION=v0.29.0
 WORKDIR /tmp
 RUN --mount=type=cache,target=/root/.cache/uv \
     git clone https://github.com/pytorch/vision.git && \
     cd vision && \
     git checkout $TORCH_VISION_VERSION && \
-    uv pip install torch==2.13.0 --index-url https://download.pytorch.org/whl/cpu && \
+    uv pip install torch==2.14.0 --index-url https://download.pytorch.org/whl/test/cpu && \
     python setup.py bdist_wheel
 
 FROM python-install AS hf-xet-builder
@@ -156,7 +156,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 # # Final build stage
 FROM python-install AS vllm-cpu
 ARG PYTHON_VERSION
-ARG PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu"
+ARG PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/test/cpu"
 # Set correct library path for torch and numactl
 ENV LD_LIBRARY_PATH="/opt/vllm/lib64/python${PYTHON_VERSION}/site-packages/torch/lib:/usr/local/lib:/opt/rh/gcc-toolset-14/root/usr/lib64:$LD_LIBRARY_PATH"
 ENV C_INCLUDE_PATH="/usr/local/include:$C_INCLUDE_PATH"

docker/versions.json

@@ -26,7 +26,7 @@
       "default": "https://bootstrap.pypa.io/get-pip.py"
     },
     "PYTORCH_CUDA_INDEX_BASE_URL": {
-      "default": "https://download.pytorch.org/whl"
+      "default": "https://download.pytorch.org/whl/test"
     },
     "PIP_KEYRING_PROVIDER": {
       "default": "disabled"

pyproject.toml

@@ -7,7 +7,7 @@ requires = [
     "setuptools>=77.0.3,<81.0.0",
     "setuptools-scm>=8.0",
     "setuptools-rust>=1.9.0",
-    "torch == 2.13.0",
+    "torch == 2.14.0",
     "wheel",
     "jinja2",
 ]

requirements/build/cpu.txt

@@ -1,11 +1,12 @@
+--extra-index-url https://download.pytorch.org/whl/test/cpu
 cmake>=3.26.1
 ninja
 packaging>=24.2
 setuptools==77.0.3 # this version can reuse CMake build dir
 setuptools-scm>=8
 setuptools-rust>=1.9.0
-torch==2.13.0+cpu; platform_machine == "x86_64" or platform_machine == "s390x" or platform_machine == "aarch64"
-torch==2.13.0; platform_system == "Darwin" or platform_machine == "ppc64le"  or platform_machine == "riscv64"
+torch==2.14.0+cpu; platform_machine == "x86_64" or platform_machine == "s390x" or platform_machine == "aarch64"
+torch==2.14.0; platform_system == "Darwin" or platform_machine == "ppc64le"  or platform_machine == "riscv64"
 wheel
 jinja2>=3.1.6
 regex

requirements/build/cuda.txt

@@ -1,11 +1,12 @@
 # Should be mirrored in pyproject.toml
+--extra-index-url https://download.pytorch.org/whl/test/cu130
 cmake>=3.26.1
 ninja
 packaging>=24.2
 setuptools>=77.0.3,<81.0.0
 setuptools-scm>=8
 setuptools-rust>=1.9.0
-torch==2.13.0
+torch==2.14.0
 wheel
 jinja2>=3.1.6
 regex

requirements/cpu.txt

@@ -1,13 +1,15 @@
 # Common dependencies
 -r common.txt
 
+--extra-index-url https://download.pytorch.org/whl/test/cpu
+
 setuptools==77.0.3 # this version can reuse CMake build dir
 
 numba == 0.65.0; platform_machine != "s390x" # Required for N-gram speculative decoding
 
 # Dependencies for CPUs
-torch==2.13.0+cpu; platform_machine == "x86_64" or platform_machine == "s390x" or platform_machine == "aarch64"
-torch==2.13.0; platform_system == "Darwin" or platform_machine == "ppc64le" or platform_machine == "riscv64"
+torch==2.14.0+cpu; platform_machine == "x86_64" or platform_machine == "s390x" or platform_machine == "aarch64"
+torch==2.14.0; platform_system == "Darwin" or platform_machine == "ppc64le" or platform_machine == "riscv64"
 
 # required for the image processor of minicpm-o-2_6, this must be updated alongside torch
 torchaudio; platform_machine != "s390x" and platform_machine != "riscv64"

requirements/cuda.txt

@@ -4,10 +4,10 @@
 numba == 0.65.0 # Required for N-gram speculative decoding
 
 # Dependencies for NVIDIA GPUs
-torch==2.13.0
+torch==2.14.0
 torchaudio==2.11.0
 # These must be updated alongside torch
-torchvision==0.28.0 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
+torchvision==0.29.0 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
 torchcodec >= 0.14
 PyNvVideoCodec==2.0.4
 # FlashInfer should be updated together with the Dockerfile

requirements/test/cpu.txt

@@ -1123,7 +1123,7 @@ tokenizers==0.23.1
     #   -r requirements/test/../common.txt
     #   cohere
     #   transformers
-torch==2.13.0+cpu
+torch==2.14.0+cpu
     # via
     #   -r requirements/test/cuda.in
     #   accelerate
@@ -1151,7 +1151,7 @@ torchaudio==2.11.0+cpu
     #   vocos
 torchcodec==0.14.0+cpu
     # via -r requirements/test/cuda.in
-torchvision==0.28.0+cpu
+torchvision==0.29.0+cpu
     # via
     #   -r requirements/test/cuda.in
     #   open-clip-torch
@@ -1186,7 +1186,7 @@ transformers==5.16.1
     #   xgrammar
 transformers-stream-generator==0.0.5
     # via -r requirements/test/cuda.in
-triton==3.6.0
+triton==3.8.0
     # via xgrammar
 tritonclient==2.64.0
     # via -r requirements/test/cuda.in

requirements/test/cuda.in

@@ -28,9 +28,9 @@ soundfile # required for audio tests
 jiwer # required for audio tests
 tblib # for pickling test exceptions
 timm >=1.0.17 # required for internvl and gemma3n-mm test
-torch==2.13.0
+torch==2.14.0
 torchaudio==2.11.0
-torchvision==0.28.0
+torchvision==0.29.0
 transformers_stream_generator # required for qwen-vl test
 matplotlib # required for qwen-vl test
 mistral_common[image,audio] >= 1.11.6 # required for voxtral test