e6e3c55ef2 · tduncan/Nvidia-Optimization-Stack-VLLM

.buildkite/release-pipeline.yaml

+6 -6

@@ -3,7 +3,7 @@ steps:

3 agents:

4 queue: cpu_queue_postmerge

5 commands:

6 - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.4.0 --tag vllm-ci:build-image --target build --progress plain ."

6 + - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.4.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."

7 - "mkdir artifacts"

8 - "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'"

9 - "bash .buildkite/upload-wheels.sh"

@@ -14,7 +14,7 @@ steps:

14 agents:

15 queue: cpu_queue_postmerge

16 commands:

17 - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.1.0 --tag vllm-ci:build-image --target build --progress plain ."

17 + - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.1.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."

18 - "mkdir artifacts"

19 - "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'"

20 - "bash .buildkite/upload-wheels.sh"

@@ -31,7 +31,7 @@ steps:

31 agents:

32 queue: cpu_queue_postmerge

33 commands:

34 - - "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=11.8.0 --tag vllm-ci:build-image --target build --progress plain ."

34 + - "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=11.8.0 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."

35 - "mkdir artifacts"

36 - "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'"

37 - "bash .buildkite/upload-wheels.sh"

@@ -48,7 +48,7 @@ steps:

48 queue: cpu_queue_postmerge

49 commands:

50 - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"

51 - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.4.0 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT --target vllm-openai --progress plain ."

51 + - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.4.0 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT --target vllm-openai --progress plain -f docker/Dockerfile ."

52 - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"

53

54 - label: "Build and publish TPU release image"

@@ -57,7 +57,7 @@ steps:

57 agents:

58 queue: tpu_queue_postmerge

59 commands:

60 - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --tag vllm/vllm-tpu:nightly --tag vllm/vllm-tpu:$BUILDKITE_COMMIT --progress plain -f Dockerfile.tpu ."

60 + - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --tag vllm/vllm-tpu:nightly --tag vllm/vllm-tpu:$BUILDKITE_COMMIT --progress plain -f docker/Dockerfile.tpu ."

61 - "docker push vllm/vllm-tpu:nightly"

62 - "docker push vllm/vllm-tpu:$BUILDKITE_COMMIT"

63 plugins:

@@ -82,7 +82,7 @@ steps:

82 queue: cpu_queue_postmerge

83 commands:

84 - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"

85 - - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f Dockerfile.cpu ."

85 + - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg GIT_REPO_CHECK=1 --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version) --tag public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:latest --progress plain --target vllm-openai -f docker/Dockerfile.cpu ."

86 - "docker push public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:$(buildkite-agent meta-data get release-version)"

87 env:

88 DOCKER_BUILDKIT: "1"

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

+1 -1

@@ -10,5 +10,5 @@ trap remove_docker_container EXIT

10 remove_docker_container

11

12 # Try building the docker image

13 -docker build -t cpu-test -f Dockerfile.ppc64le .

13 +docker build -t cpu-test -f docker/Dockerfile.ppc64le .

14

.buildkite/run-cpu-test.sh

+2 -2

@@ -18,8 +18,8 @@ trap remove_docker_container EXIT

18 remove_docker_container

19

20 # Try building the docker image

21 -numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --tag cpu-test-"$BUILDKITE_BUILD_NUMBER" --target vllm-test -f Dockerfile.cpu .

22 -numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" --tag cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2 --target vllm-test -f Dockerfile.cpu .

21 +numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --tag cpu-test-"$BUILDKITE_BUILD_NUMBER" --target vllm-test -f docker/Dockerfile.cpu .

22 +numactl -C "$CORE_RANGE" -N "$NUMA_NODE" docker build --build-arg VLLM_CPU_DISABLE_AVX512="true" --tag cpu-test-"$BUILDKITE_BUILD_NUMBER"-avx2 --target vllm-test -f docker/Dockerfile.cpu .

23

24 # Run the image, setting --shm-size=4g for tensor parallel.

25 docker run -itd --entrypoint /bin/bash -v ~/.cache/huggingface:/root/.cache/huggingface --cpuset-cpus="$CORE_RANGE" \

.buildkite/run-gh200-test.sh

+1

@@ -9,6 +9,7 @@ python3 use_existing_torch.py

9

10 # Try building the docker image

11 DOCKER_BUILDKIT=1 docker build . \

12 + --file docker/Dockerfile \

13 --target vllm-openai \

14 --platform "linux/arm64" \

15 -t gh200-test \

.buildkite/run-hpu-test.sh

+1 -1

@@ -5,7 +5,7 @@

5 set -ex

6

7 # Try building the docker image

8 -docker build -t hpu-test-env -f Dockerfile.hpu .

8 +docker build -t hpu-test-env -f docker/Dockerfile.hpu .

9

10 # Setup cleanup

11 # certain versions of HPU software stack have a bug that can

.buildkite/run-neuron-test.sh

+1 -1

@@ -35,7 +35,7 @@ else

35 date "+%s" > /tmp/neuron-docker-build-timestamp

36 fi

37

38 -docker build -t "${image_name}" -f Dockerfile.neuron .

38 +docker build -t "${image_name}" -f docker/Dockerfile.neuron .

39

40 # Setup cleanup

41 remove_docker_container() {

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

+1 -1

@@ -3,7 +3,7 @@

3 set -e

4

5 # Build the docker image.

6 -docker build -f Dockerfile.tpu -t vllm-tpu .

6 +docker build -f docker/Dockerfile.tpu -t vllm-tpu .

7

8 # Set up cleanup.

9 remove_docker_container() { docker rm -f tpu-test || true; }

.buildkite/run-xpu-test.sh

+1 -1

@@ -8,7 +8,7 @@ image_name="xpu/vllm-ci:${BUILDKITE_COMMIT}"

8 container_name="xpu_${BUILDKITE_COMMIT}_$(tr -dc A-Za-z0-9 < /dev/urandom | head -c 10; echo)"

9

10 # Try building the docker image

11 -docker build -t ${image_name} -f Dockerfile.xpu .

11 +docker build -t ${image_name} -f docker/Dockerfile.xpu .

12

13 # Setup cleanup

14 remove_docker_container() {

.github/mergify.yml

+1 -1

@@ -19,7 +19,7 @@ pull_request_rules:

19 - files~=\.buildkite/

20 - files~=^cmake/

21 - files=CMakeLists.txt

22 - - files~=^Dockerfile

22 + - files~=^docker/Dockerfile

23 - files~=^requirements.*\.txt

24 - files=setup.py

25 actions:

.github/workflows/lint-and-deploy.yaml

+1 -1

@@ -50,7 +50,7 @@ jobs:

50 uses: helm/kind-action@a1b0e391336a6ee6713a0583f8c6240d70863de3 # v1.12.0

51

52 - name: Build the Docker image vllm cpu

53 - run: docker buildx build -f Dockerfile.cpu -t vllm-cpu-env .

53 + run: docker buildx build -f docker/Dockerfile.cpu -t vllm-cpu-env .

54

55 - name: Configuration of docker images, network and namespace for the kind cluster

56 run: |

CMakeLists.txt

+1 -1

@@ -44,7 +44,7 @@ set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1

44 #

45 # Note: the CUDA torch version is derived from pyproject.toml and various

46 # requirements.txt files and should be kept consistent. The ROCm torch

47 -# versions are derived from Dockerfile.rocm

47 +# versions are derived from docker/Dockerfile.rocm

48 #

49 set(TORCH_SUPPORTED_VERSION_CUDA "2.6.0")

50 set(TORCH_SUPPORTED_VERSION_ROCM "2.6.0")

docker/Dockerfile renamed

docker/Dockerfile.arm renamed

docker/Dockerfile.cpu renamed

docker/Dockerfile.hpu renamed

docker/Dockerfile.neuron renamed

docker/Dockerfile.ppc64le renamed

docker/Dockerfile.rocm renamed

docker/Dockerfile.rocm_base renamed

docker/Dockerfile.s390x renamed

docker/Dockerfile.tpu renamed

docker/Dockerfile.xpu renamed

docs/source/contributing/dockerfile/dockerfile.md

+3 -3

@@ -1,6 +1,6 @@

1 # Dockerfile

2

3 -We provide a <gh-file:Dockerfile> to construct the image for running an OpenAI compatible server with vLLM.

3 +We provide a <gh-file:docker/Dockerfile> to construct the image for running an OpenAI compatible server with vLLM.

4 More information about deploying with Docker can be found [here](#deployment-docker).

5

6 Below is a visual representation of the multi-stage Dockerfile. The build graph contains the following nodes:

@@ -28,7 +28,7 @@ The edges of the build graph represent:

28 > Commands to regenerate the build graph (make sure to run it **from the \`root\` directory of the vLLM repository** where the dockerfile is present):

29 >

30 > ```bash

31 - > dockerfilegraph -o png --legend --dpi 200 --max-label-length 50 --filename Dockerfile

31 + > dockerfilegraph -o png --legend --dpi 200 --max-label-length 50 --filename docker/Dockerfile

32 > ```

33 >

34 > or in case you want to run it directly with the docker image:

@@ -43,7 +43,7 @@ The edges of the build graph represent:

43 > --output png \

44 > --dpi 200 \

45 > --max-label-length 50 \

46 - > --filename Dockerfile \

46 + > --filename docker/Dockerfile \

47 > --legend

48 > ```

49 >

docs/source/contributing/overview.md

+1 -1

@@ -45,7 +45,7 @@ pytest tests/

45 ```

46

47 :::{tip}

48 -Since the <gh-file:Dockerfile> ships with Python 3.12, all tests in CI (except `mypy`) are run with Python 3.12.

48 +Since the <gh-file:docker/Dockerfile> ships with Python 3.12, all tests in CI (except `mypy`) are run with Python 3.12.

49

50 Therefore, we recommend developing with Python 3.12 to minimise the chance of your local environment clashing with our CI environment.

51 :::

docs/source/deployment/docker.md

+3 -2

@@ -61,11 +61,11 @@ RUN uv pip install --system git+https://github.com/huggingface/transformers.git

61

62 ## Building vLLM's Docker Image from Source

63

64 -You can build and run vLLM from source via the provided <gh-file:Dockerfile>. To build vLLM:

64 +You can build and run vLLM from source via the provided <gh-file:docker/Dockerfile>. To build vLLM:

65

66 ```console

67 # optionally specifies: --build-arg max_jobs=8 --build-arg nvcc_threads=2

68 -DOCKER_BUILDKIT=1 docker build . --target vllm-openai --tag vllm/vllm-openai

68 +DOCKER_BUILDKIT=1 docker build . --target vllm-openai --tag vllm/vllm-openai --file docker/Dockerfile

69 ```

70

71 :::{note}

@@ -92,6 +92,7 @@ Keep an eye on memory usage with parallel jobs as it can be substantial (see exa

92 # Example of building on Nvidia GH200 server. (Memory usage: ~15GB, Build time: ~1475s / ~25 min, Image size: 6.93GB)

93 $ python3 use_existing_torch.py

94 $ DOCKER_BUILDKIT=1 docker build . \

95 + --file docker/Dockerfile \

96 --target vllm-openai \

97 --platform "linux/arm64" \

98 -t vllm/vllm-gh200-openai:latest \

docs/source/deployment/nginx.md

+2 -2

@@ -69,14 +69,14 @@ server {

69

70 ```console

71 cd $vllm_root

72 -docker build -f Dockerfile . --tag vllm

72 +docker build -f docker/Dockerfile . --tag vllm

73 ```

74

75 If you are behind proxy, you can pass the proxy settings to the docker build command as shown below:

76

77 ```console

78 cd $vllm_root

79 -docker build -f Dockerfile . --tag vllm --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_proxy

79 +docker build -f docker/Dockerfile . --tag vllm --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_proxy

80 ```

81

82 (nginxloadbalancer-nginx-docker-network)=

docs/source/getting_started/installation/ai_accelerator/hpu-gaudi.inc.md

+1 -1

@@ -86,7 +86,7 @@ Currently, there are no pre-built Intel Gaudi images.

86 ### Build image from source

87

88 ```console

89 -docker build -f Dockerfile.hpu -t vllm-hpu-env .

89 +docker build -f docker/Dockerfile.hpu -t vllm-hpu-env .

90 docker run -it --runtime=habana -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --net=host --rm vllm-hpu-env

91 ```

92

docs/source/getting_started/installation/ai_accelerator/neuron.inc.md

+1 -1

@@ -132,7 +132,7 @@ Currently, there are no pre-built Neuron images.

132

133 See <project:#deployment-docker-build-image-from-source> for instructions on building the Docker image.

134

135 -Make sure to use <gh-file:Dockerfile.neuron> in place of the default Dockerfile.

135 +Make sure to use <gh-file:docker/Dockerfile.neuron> in place of the default Dockerfile.

136

137 ## Extra information

138

docs/source/getting_started/installation/ai_accelerator/tpu.inc.md

+2 -2

@@ -169,10 +169,10 @@ See <project:#deployment-docker-pre-built-image> for instructions on using the o

169

170 ### Build image from source

171

172 -You can use <gh-file:Dockerfile.tpu> to build a Docker image with TPU support.

172 +You can use <gh-file:docker/Dockerfile.tpu> to build a Docker image with TPU support.

173

174 ```console

175 -docker build -f Dockerfile.tpu -t vllm-tpu .

175 +docker build -f docker/Dockerfile.tpu -t vllm-tpu .

176 ```

177

178 Run the Docker image with the following command:

docs/source/getting_started/installation/cpu.md

+3 -3

@@ -177,7 +177,7 @@ Currently, there are no pre-built CPU wheels.

177 ### Build image from source

178

179 ```console

180 -$ docker build -f Dockerfile.cpu --tag vllm-cpu-env --target vllm-openai .

180 +$ docker build -f docker/Dockerfile.cpu --tag vllm-cpu-env --target vllm-openai .

181

182 # Launching OpenAI server

183 $ docker run --rm \

@@ -193,11 +193,11 @@ $ docker run --rm \

193 ```

194

195 ::::{tip}

196 -For ARM or Apple silicon, use `Dockerfile.arm`

196 +For ARM or Apple silicon, use `docker/Dockerfile.arm`

197 ::::

198

199 ::::{tip}

200 -For IBM Z (s390x), use `Dockerfile.s390x` and in `docker run` use flag `--dtype float`

200 +For IBM Z (s390x), use `docker/Dockerfile.s390x` and in `docker run` use flag `--dtype float`

201 ::::

202

203 ## Supported features

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

+7 -7

@@ -123,7 +123,7 @@ Building the Docker image from source is the recommended way to use vLLM with RO

123

124 #### (Optional) Build an image with ROCm software stack

125

126 -Build a docker image from <gh-file:Dockerfile.rocm_base> which setup ROCm software stack needed by the vLLM.

126 +Build a docker image from <gh-file:docker/Dockerfile.rocm_base> which setup ROCm software stack needed by the vLLM.

127 **This step is optional as this rocm_base image is usually prebuilt and store at [Docker Hub](https://hub.docker.com/r/rocm/vllm-dev) under tag `rocm/vllm-dev:base` to speed up user experience.**

128 If you choose to build this rocm_base image yourself, the steps are as follows.

129

@@ -140,12 +140,12 @@ It is important that the user kicks off the docker build using buildkit. Either

140 To build vllm on ROCm 6.3 for MI200 and MI300 series, you can use the default:

141

142 ```console

143 -DOCKER_BUILDKIT=1 docker build -f Dockerfile.rocm_base -t rocm/vllm-dev:base .

143 +DOCKER_BUILDKIT=1 docker build -f docker/Dockerfile.rocm_base -t rocm/vllm-dev:base .

144 ```

145

146 #### Build an image with vLLM

147

148 -First, build a docker image from <gh-file:Dockerfile.rocm> and launch a docker container from the image.

148 +First, build a docker image from <gh-file:docker/Dockerfile.rocm> and launch a docker container from the image.

149 It is important that the user kicks off the docker build using buildkit. Either the user put `DOCKER_BUILDKIT=1` as environment variable when calling docker build command, or the user needs to setup buildkit in the docker daemon configuration /etc/docker/daemon.json as follows and restart the daemon:

150

151 ```console

@@ -156,10 +156,10 @@ It is important that the user kicks off the docker build using buildkit. Either

156 }

157 ```

158

159 -<gh-file:Dockerfile.rocm> uses ROCm 6.3 by default, but also supports ROCm 5.7, 6.0, 6.1, and 6.2, in older vLLM branches.

159 +<gh-file:docker/Dockerfile.rocm> uses ROCm 6.3 by default, but also supports ROCm 5.7, 6.0, 6.1, and 6.2, in older vLLM branches.

160 It provides flexibility to customize the build of docker image using the following arguments:

161

162 -- `BASE_IMAGE`: specifies the base image used when running `docker build`. The default value `rocm/vllm-dev:base` is an image published and maintained by AMD. It is being built using <gh-file:Dockerfile.rocm_base>

162 +- `BASE_IMAGE`: specifies the base image used when running `docker build`. The default value `rocm/vllm-dev:base` is an image published and maintained by AMD. It is being built using <gh-file:docker/Dockerfile.rocm_base>

163 - `USE_CYTHON`: An option to run cython compilation on a subset of python files upon docker build

164 - `BUILD_RPD`: Include RocmProfileData profiling tool in the image

165 - `ARG_PYTORCH_ROCM_ARCH`: Allows to override the gfx architecture values from the base docker image

@@ -169,13 +169,13 @@ Their values can be passed in when running `docker build` with `--build-arg` opt

169 To build vllm on ROCm 6.3 for MI200 and MI300 series, you can use the default:

170

171 ```console

172 -DOCKER_BUILDKIT=1 docker build -f Dockerfile.rocm -t vllm-rocm .

172 +DOCKER_BUILDKIT=1 docker build -f docker/Dockerfile.rocm -t vllm-rocm .

173 ```

174

175 To build vllm on ROCm 6.3 for Radeon RX7900 series (gfx1100), you should pick the alternative base image:

176

177 ```console

178 -DOCKER_BUILDKIT=1 docker build --build-arg BASE_IMAGE="rocm/vllm-dev:navi_base" -f Dockerfile.rocm -t vllm-rocm .

178 +DOCKER_BUILDKIT=1 docker build --build-arg BASE_IMAGE="rocm/vllm-dev:navi_base" -f docker/Dockerfile.rocm -t vllm-rocm .

179 ```

180

181 To run the above docker image `vllm-rocm`, use the below command:

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

+1 -1

@@ -54,7 +54,7 @@ Currently, there are no pre-built XPU images.

54 ### Build image from source

55

56 ```console

57 -$ docker build -f Dockerfile.xpu -t vllm-xpu-env --shm-size=4g .

57 +$ docker build -f docker/Dockerfile.xpu -t vllm-xpu-env --shm-size=4g .

58 $ docker run -it \

59 --rm \

60 --network=host \

docs/source/getting_started/quickstart.md

+1 -1

@@ -208,5 +208,5 @@ Currently, vLLM supports multiple backends for efficient Attention computation a

208 If desired, you can also manually set the backend of your choice by configuring the environment variable `VLLM_ATTENTION_BACKEND` to one of the following options: `FLASH_ATTN`, `FLASHINFER` or `XFORMERS`.

209

210 ```{attention}

211 -There are no pre-built vllm wheels containing Flash Infer, so you must install it in your environment first. Refer to the [Flash Infer official docs](https://docs.flashinfer.ai/) or see [Dockerfile](https://github.com/vllm-project/vllm/blob/main/Dockerfile) for instructions on how to install it.

211 +There are no pre-built vllm wheels containing Flash Infer, so you must install it in your environment first. Refer to the [Flash Infer official docs](https://docs.flashinfer.ai/) or see <gh-file:docker/Dockerfile> for instructions on how to install it.

212 ```

vllm/config.py

+2 -2

@@ -317,8 +317,8 @@ class ModelConfig:

317 ) and backend == "FLASHINFER" and find_spec("flashinfer") is None:

318 raise ValueError(

319 "VLLM_ATTENTION_BACKEND is set to FLASHINFER, but flashinfer "

320 - "module was not found."

321 - "See https://github.com/vllm-project/vllm/blob/main/Dockerfile "

320 + "module was not found. See "

321 + "https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile " # noqa: E501

322 "for instructions on how to install it.")

323

324 # The tokenizer version is consistent with the model version by default.

Move dockerfiles into their own directory (#14549)