A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 4, 16:15 UTC
[Build] Bump CUDA to 13.0.2 to match PyTorch 2.11.0 (#40669)
Signed-off-by: Dmitry Tokarev <dtokarev@nvidia.com>
Dmitry Tokarev committed
Apr 24, 2026 at 06:27 UTC
6dec49f27ece339c59d5eb92f33120c11c0c3b74
6 files changed
+13
-13
.buildkite/image_build/image_build_torch_nightly.sh
+1
-1
@@ -46,7 +46,7 @@ echo "Image not found, proceeding with build..."
46
47
# --- CUDA 13.0 for nightly builds ---
48
# Nightly CI uses CUDA 13.0 while regular CI stays on CUDA 12.9
49
-NIGHTLY_CUDA_VERSION="13.0.0"
49
+NIGHTLY_CUDA_VERSION="13.0.2"
50
NIGHTLY_BUILD_BASE_IMAGE="nvidia/cuda:${NIGHTLY_CUDA_VERSION}-devel-ubuntu22.04"
51
NIGHTLY_FINAL_BASE_IMAGE="nvidia/cuda:${NIGHTLY_CUDA_VERSION}-base-ubuntu22.04"
52
.buildkite/release-pipeline.yaml
+6
-6
@@ -37,7 +37,7 @@ steps:
37
agents:
38
queue: arm64_cpu_queue_release
39
commands:
40
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
40
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
41
- "mkdir artifacts"
42
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
43
- "bash .buildkite/scripts/upload-nightly-wheels.sh manylinux_2_35"
@@ -76,7 +76,7 @@ steps:
76
agents:
77
queue: cpu_queue_release
78
commands:
79
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
79
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
80
- "mkdir artifacts"
81
- "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
82
- "bash .buildkite/scripts/upload-nightly-wheels.sh manylinux_2_35"
@@ -121,7 +121,7 @@ steps:
121
queue: cpu_queue_release
122
commands:
123
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
124
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
124
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
125
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
126
# re-tag to default image tag and push, just in case arm64 build fails
127
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
@@ -134,7 +134,7 @@ steps:
134
queue: arm64_cpu_queue_release
135
commands:
136
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
137
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
137
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
138
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
139
140
- label: "Build release image - x86_64 - CUDA 12.9"
@@ -167,7 +167,7 @@ steps:
167
queue: cpu_queue_release
168
commands:
169
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
170
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
170
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
171
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404"
172
- "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-ubuntu2404"
173
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-ubuntu2404"
@@ -179,7 +179,7 @@ steps:
179
queue: arm64_cpu_queue_release
180
commands:
181
- "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
182
- - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
182
+ - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
183
- "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404"
184
185
- label: "Build release image - x86_64 - CUDA 12.9 - Ubuntu 24.04"
docker/Dockerfile
+1
-1
@@ -22,7 +22,7 @@
22
# docker buildx bake -f docker/docker-bake.hcl -f docker/versions.json
23
# =============================================================================
24
25
-ARG CUDA_VERSION=13.0.0
25
+ARG CUDA_VERSION=13.0.2
26
ARG PYTHON_VERSION=3.12
27
ARG UBUNTU_VERSION=22.04
28
docker/versions.json
+3
-3
@@ -2,7 +2,7 @@
2
"_comment": "Auto-generated from Dockerfile ARGs. Do not edit manually. Run: python tools/generate_versions_json.py",
3
"variable": {
4
"CUDA_VERSION": {
5
- "default": "13.0.0"
5
+ "default": "13.0.2"
6
},
7
"PYTHON_VERSION": {
8
"default": "3.12"
@@ -11,10 +11,10 @@
11
"default": "22.04"
12
},
13
"BUILD_BASE_IMAGE": {
14
- "default": "nvidia/cuda:13.0.0-devel-ubuntu22.04"
14
+ "default": "nvidia/cuda:13.0.2-devel-ubuntu22.04"
15
},
16
"FINAL_BASE_IMAGE": {
17
- "default": "nvidia/cuda:13.0.0-base-ubuntu22.04"
17
+ "default": "nvidia/cuda:13.0.2-base-ubuntu22.04"
18
},
19
"GET_PIP_URL": {
20
"default": "https://bootstrap.pypa.io/get-pip.py"
docs/assets/contributing/dockerfile-stages-dependency.png
Binary files a/docs/assets/contributing/dockerfile-stages-dependency.png and b/docs/assets/contributing/dockerfile-stages-dependency.png differ
docs/getting_started/installation/gpu.cuda.inc.md
+2
-2
@@ -375,8 +375,8 @@ For (G)B300, we recommend using CUDA 13, as shown in the following command.
375
376
```bash
377
DOCKER_BUILDKIT=1 docker build \
378
- --build-arg CUDA_VERSION=13.0.1 \
379
- --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 \
378
+ --build-arg CUDA_VERSION=13.0.2 \
379
+ --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 \
380
--build-arg max_jobs=256 \
381
--build-arg nvcc_threads=2 \
382
--build-arg RUN_WHEEL_CHECK=false \