gpustack-runner 0.1.24.post2__tar.gz → 0.1.24.post4__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/PKG-INFO +21 -21
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/README.md +20 -20
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/_version.py +2 -2
- gpustack_runner-0.1.24.post4/gpustack_runner/_version_appendix.py +1 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/runner.py.json +161 -7
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_sglang_reinstall_kernel/cann/Dockerfile +74 -0
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_sglang_reinstall_kernel/matrix.yaml +28 -0
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_vllm_patch_dp/cuda/Dockerfile +25 -0
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_vllm_patch_dp/cuda/patches/vllm_001_wrong_dp_ray.patch +41 -0
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_vllm_patch_dp/matrix.yaml +55 -0
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_vllm_patch_dp/rocm/Dockerfile +25 -0
- gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_vllm_patch_dp/rocm/patches/vllm_001_wrong_dp_ray.patch +41 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/README.md +2 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/cann/Dockerfile +31 -3
- gpustack_runner-0.1.24.post4/pack/cann/patches/vllm/001_wrong_dp_ray.patch +41 -0
- gpustack_runner-0.1.24.post4/pack/cann/patches/vllm_omni/001_wrong_patch.patch +13 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/cuda/Dockerfile +35 -10
- gpustack_runner-0.1.24.post4/pack/cuda/patches/vllm/001_wrong_dp_ray.patch +41 -0
- gpustack_runner-0.1.24.post4/pack/cuda/patches/vllm_omni/001_wrong_patch.patch +13 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/matrix.yaml +17 -17
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/rocm/Dockerfile +36 -10
- gpustack_runner-0.1.24.post4/pack/rocm/patches/vllm/001_wrong_dp_ray.patch +41 -0
- gpustack_runner-0.1.24.post4/pack/rocm/patches/vllm_omni/001_wrong_patch.patch +13 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_list_runners_by_backend.json +161 -7
- gpustack_runner-0.1.24.post2/gpustack_runner/_version_appendix.py +0 -1
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/.codespelldict +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/.codespellrc +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/.gitattributes +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/.gitignore +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/.pre-commit-config.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/.python-version +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/LICENSE +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/Makefile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/docs/index.md +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/docs/modules/gpustack_runner.md +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/__init__.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/__main__.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/__utils__.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/_version.pyi +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/cmds/__init__.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/cmds/__types__.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/cmds/images.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/envs.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/runner.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/hatch.toml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/mkdocs.yml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251020_vllm_install_lmcache/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251020_vllm_install_lmcache/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251020_vllm_install_lmcache/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251020_vllm_install_lmcache/rocm/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_client/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_client/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_client/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_client/rocm/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_default/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_default/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251022_vllm_install_ray_default/rocm/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251024_vllm_install_nvidia_hpcx/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251024_vllm_install_nvidia_hpcx/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251024_vllm_reinstall_lmcache/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251024_vllm_reinstall_lmcache/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251029_vllm_reinstall_ray/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251029_vllm_reinstall_ray/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251103_mindie_refresh_entrypoint/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251103_mindie_refresh_entrypoint/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251105_vllm_polish_nvidia_hpcx/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251105_vllm_polish_nvidia_hpcx/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251106_vllm_install_ep_kernel/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251106_vllm_install_ep_kernel/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251107_vllm_reinstall_lmcache/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251107_vllm_reinstall_lmcache/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251110_sglang_install_diffusion/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251110_sglang_install_diffusion/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251110_sglang_install_flashattn/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251110_sglang_install_flashattn/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251125_mindie_install_posix_ipc/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251125_mindie_install_posix_ipc/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251201_vllm_patch_qwen2_5_vl/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251201_vllm_patch_qwen2_5_vl/cuda/patches/vllm_001_disable_flashatten_in_qwen2_5_vl.patch +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251201_vllm_patch_qwen2_5_vl/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251209_mindie_install_av/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251209_mindie_install_av/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251213_mindie_patch_minicpm_qwen2_v2/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251213_mindie_patch_minicpm_qwen2_v2/cann/patches.zip +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251213_mindie_patch_minicpm_qwen2_v2/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251213_sglang_patch_server_args/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251213_sglang_patch_server_args/cuda/patches/sglang_001_fix_server_args.patch +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251213_sglang_patch_server_args/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251214_cuda_several_patches/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251214_cuda_several_patches/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251215_cann_several_patches/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251215_cann_several_patches/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251216_sglang_uninstall_runai_model_streamer/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251216_sglang_uninstall_runai_model_streamer/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251219_rocm_install_petit_kernel/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251219_rocm_install_petit_kernel/rocm/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251219_vllm_install_audio_extra/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251219_vllm_install_audio_extra/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251219_vllm_install_audio_extra/rocm/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251224_mindie_patch_atb_config/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20251224_mindie_patch_atb_config/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20260105_vllm_install_omni/cann/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20260105_vllm_install_omni/cuda/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20260105_vllm_install_omni/matrix.yaml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/.post_operation/20260105_vllm_install_omni/rocm/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/cann/mindie-atb-models_2.3.0_linux-amd64_py3.11_torch2.1.0-abi0.tar.gz +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/cann/mindie-atb-models_2.3.0_linux-arm64_py3.11_torch2.1.0-abi0.tar.gz +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/cann/patches/mindie.zip +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/corex/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/discard_runner.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/dtk/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/expand_matrix.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/hggc/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/maca/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/merge_runner.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/musa/Dockerfile +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/prune_runner.sh +0 -0
- /gpustack_runner-0.1.24.post2/pack/rocm/patches/sglang_001_wrong_vram.patch → /gpustack_runner-0.1.24.post4/pack/rocm/patches/sglang/001_wrong_vram.patch +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/squash_expand_matrix.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pack/squash_image.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pyproject.toml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/pytest.ini +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/ruff.toml +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/__init__.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_docker_image.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_list_backend_runners.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_list_runners_by_prefix.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_list_service_runners.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_merge_image.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_replace_image_with.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_split_image.json +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/test_runner.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/test_utils.py +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/activate +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat_tool_current_date_time.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat_tool_get_temperature.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat_tool_get_weather.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat_tool_square_of_number.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat_tool_square_root_of_number.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/chat_tool_where_am_i.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/run_runner.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/tools/run_runner_cluster.sh +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/uv.lock +0 -0
- {gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/uv.toml +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: gpustack-runner
|
|
3
|
-
Version: 0.1.24.
|
|
3
|
+
Version: 0.1.24.post4
|
|
4
4
|
Summary: GPUStack Runner is library for registering runnable accelerated backends and services in GPUStack.
|
|
5
5
|
Project-URL: Homepage, https://github.com/gpustack/runner
|
|
6
6
|
Project-URL: Bug Tracker, https://github.com/gpustack/gpustack/issues
|
|
@@ -52,17 +52,17 @@ The following table lists the supported accelerated backends and their correspon
|
|
|
52
52
|
vllm-ascend [#3316](https://github.com/vllm-project/vllm-ascend/issues/3316)
|
|
53
53
|
and [#2795](https://github.com/vllm-project/vllm-ascend/issues/2795).
|
|
54
54
|
|
|
55
|
-
| CANN Version <br/> (Variant) | MindIE | vLLM
|
|
56
|
-
|
|
57
|
-
| 8.5 (A3/910C) | `2.3.0` | `0.14.1`, `0.13.0`
|
|
58
|
-
| 8.5 (910B) | `2.3.0` | `0.14.1`, `0.13.0`
|
|
59
|
-
| 8.5 (310P) | `2.3.0` | `0.14.1`
|
|
60
|
-
| 8.3 (A3/910C) | `2.2.rc1` | `0.12.0`, `0.11.0`
|
|
61
|
-
| 8.3 (910B) | `2.2.rc1` | `0.12.0`, `0.11.0`
|
|
62
|
-
| 8.3 (310P) | `2.2.rc1` |
|
|
63
|
-
| 8.2 (A3/910C) | `2.1.rc2` | `0.10.2`,
|
|
64
|
-
| 8.2 (910B) | `2.1.rc2` | `0.10.2`,
|
|
65
|
-
| 8.2 (310P) | `2.1.rc2` | `0.10.0`, `0.9.2`
|
|
55
|
+
| CANN Version <br/> (Variant) | MindIE | vLLM | SGLang |
|
|
56
|
+
|------------------------------|-----------|--------------------------------------------------------------------|------------------------|
|
|
57
|
+
| 8.5 (A3/910C) | `2.3.0` | `0.14.1`, `0.13.0` | `0.5.8` |
|
|
58
|
+
| 8.5 (910B) | `2.3.0` | `0.14.1`, `0.13.0` | `0.5.8` |
|
|
59
|
+
| 8.5 (310P) | `2.3.0` | `0.14.1` | |
|
|
60
|
+
| 8.3 (A3/910C) | `2.2.rc1` | `0.12.0`, `0.11.0` | `0.5.7`, `0.5.6.post2` |
|
|
61
|
+
| 8.3 (910B) | `2.2.rc1` | `0.12.0`, `0.11.0` | `0.5.7`, `0.5.6.post2` |
|
|
62
|
+
| 8.3 (310P) | `2.2.rc1` | | |
|
|
63
|
+
| 8.2 (A3/910C) | `2.1.rc2` | `0.10.2`, ~~`0.10.1.1`~~ | `0.5.2`, `0.5.1.post3` |
|
|
64
|
+
| 8.2 (910B) | `2.1.rc2` | `0.10.2`, ~~`0.10.1.1`~~, <br/>`0.10.0`, `0.9.2`, <br/>~~`0.9.1`~~ | `0.5.2`, `0.5.1.post3` |
|
|
65
|
+
| 8.2 (310P) | `2.1.rc2` | `0.10.0`, `0.9.2` | |
|
|
66
66
|
|
|
67
67
|
### Iluvatar CoreX
|
|
68
68
|
|
|
@@ -80,11 +80,11 @@ The following table lists the supported accelerated backends and their correspon
|
|
|
80
80
|
> - CUDA 12.6/12.4 supports Compute Capabilities:
|
|
81
81
|
`7.5 8.0+PTX 8.9 9.0+PTX`.
|
|
82
82
|
|
|
83
|
-
| CUDA Version <br/> (Variant) | vLLM
|
|
84
|
-
|
|
85
|
-
| 12.9 | `0.
|
|
86
|
-
| 12.8 | `0.
|
|
87
|
-
| 12.6 | `0.
|
|
83
|
+
| CUDA Version <br/> (Variant) | vLLM | SGLang | VoxBox |
|
|
84
|
+
|------------------------------|----------------------------------------------------------------|-----------------------------------------------------------------------------|----------|
|
|
85
|
+
| 12.9 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2` | |
|
|
86
|
+
| 12.8 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, <br/>`0.10.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`, `0.5.5.post3`, <br/>~~`0.5.4.post3`~~ | `0.0.21` |
|
|
87
|
+
| 12.6 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, <br/>`0.10.2` | | `0.0.21` |
|
|
88
88
|
|
|
89
89
|
### Hygon DTK
|
|
90
90
|
|
|
@@ -128,10 +128,10 @@ The following table lists the supported accelerated backends and their correspon
|
|
|
128
128
|
> - ROCm 6.4 SGLang supports `gfx942` only.
|
|
129
129
|
> - ROCm 7.0 SGLang supports `gfx950` only.
|
|
130
130
|
|
|
131
|
-
| ROCm Version <br/> (Variant) | vLLM
|
|
132
|
-
|
|
133
|
-
| 7.0 | `0.
|
|
134
|
-
| 6.4 | `0.
|
|
131
|
+
| ROCm Version <br/> (Variant) | vLLM | SGLang |
|
|
132
|
+
|------------------------------|-----------------------------------------------------------|-----------------------------------------------------|
|
|
133
|
+
| 7.0 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2` |
|
|
134
|
+
| 6.4 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, `0.10.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`, `0.5.5.post3` |
|
|
135
135
|
|
|
136
136
|
## Directory Structure
|
|
137
137
|
|
|
@@ -32,17 +32,17 @@ The following table lists the supported accelerated backends and their correspon
|
|
|
32
32
|
vllm-ascend [#3316](https://github.com/vllm-project/vllm-ascend/issues/3316)
|
|
33
33
|
and [#2795](https://github.com/vllm-project/vllm-ascend/issues/2795).
|
|
34
34
|
|
|
35
|
-
| CANN Version <br/> (Variant) | MindIE | vLLM
|
|
36
|
-
|
|
37
|
-
| 8.5 (A3/910C) | `2.3.0` | `0.14.1`, `0.13.0`
|
|
38
|
-
| 8.5 (910B) | `2.3.0` | `0.14.1`, `0.13.0`
|
|
39
|
-
| 8.5 (310P) | `2.3.0` | `0.14.1`
|
|
40
|
-
| 8.3 (A3/910C) | `2.2.rc1` | `0.12.0`, `0.11.0`
|
|
41
|
-
| 8.3 (910B) | `2.2.rc1` | `0.12.0`, `0.11.0`
|
|
42
|
-
| 8.3 (310P) | `2.2.rc1` |
|
|
43
|
-
| 8.2 (A3/910C) | `2.1.rc2` | `0.10.2`,
|
|
44
|
-
| 8.2 (910B) | `2.1.rc2` | `0.10.2`,
|
|
45
|
-
| 8.2 (310P) | `2.1.rc2` | `0.10.0`, `0.9.2`
|
|
35
|
+
| CANN Version <br/> (Variant) | MindIE | vLLM | SGLang |
|
|
36
|
+
|------------------------------|-----------|--------------------------------------------------------------------|------------------------|
|
|
37
|
+
| 8.5 (A3/910C) | `2.3.0` | `0.14.1`, `0.13.0` | `0.5.8` |
|
|
38
|
+
| 8.5 (910B) | `2.3.0` | `0.14.1`, `0.13.0` | `0.5.8` |
|
|
39
|
+
| 8.5 (310P) | `2.3.0` | `0.14.1` | |
|
|
40
|
+
| 8.3 (A3/910C) | `2.2.rc1` | `0.12.0`, `0.11.0` | `0.5.7`, `0.5.6.post2` |
|
|
41
|
+
| 8.3 (910B) | `2.2.rc1` | `0.12.0`, `0.11.0` | `0.5.7`, `0.5.6.post2` |
|
|
42
|
+
| 8.3 (310P) | `2.2.rc1` | | |
|
|
43
|
+
| 8.2 (A3/910C) | `2.1.rc2` | `0.10.2`, ~~`0.10.1.1`~~ | `0.5.2`, `0.5.1.post3` |
|
|
44
|
+
| 8.2 (910B) | `2.1.rc2` | `0.10.2`, ~~`0.10.1.1`~~, <br/>`0.10.0`, `0.9.2`, <br/>~~`0.9.1`~~ | `0.5.2`, `0.5.1.post3` |
|
|
45
|
+
| 8.2 (310P) | `2.1.rc2` | `0.10.0`, `0.9.2` | |
|
|
46
46
|
|
|
47
47
|
### Iluvatar CoreX
|
|
48
48
|
|
|
@@ -60,11 +60,11 @@ The following table lists the supported accelerated backends and their correspon
|
|
|
60
60
|
> - CUDA 12.6/12.4 supports Compute Capabilities:
|
|
61
61
|
`7.5 8.0+PTX 8.9 9.0+PTX`.
|
|
62
62
|
|
|
63
|
-
| CUDA Version <br/> (Variant) | vLLM
|
|
64
|
-
|
|
65
|
-
| 12.9 | `0.
|
|
66
|
-
| 12.8 | `0.
|
|
67
|
-
| 12.6 | `0.
|
|
63
|
+
| CUDA Version <br/> (Variant) | vLLM | SGLang | VoxBox |
|
|
64
|
+
|------------------------------|----------------------------------------------------------------|-----------------------------------------------------------------------------|----------|
|
|
65
|
+
| 12.9 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2` | |
|
|
66
|
+
| 12.8 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, <br/>`0.10.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`, `0.5.5.post3`, <br/>~~`0.5.4.post3`~~ | `0.0.21` |
|
|
67
|
+
| 12.6 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, <br/>`0.10.2` | | `0.0.21` |
|
|
68
68
|
|
|
69
69
|
### Hygon DTK
|
|
70
70
|
|
|
@@ -108,10 +108,10 @@ The following table lists the supported accelerated backends and their correspon
|
|
|
108
108
|
> - ROCm 6.4 SGLang supports `gfx942` only.
|
|
109
109
|
> - ROCm 7.0 SGLang supports `gfx950` only.
|
|
110
110
|
|
|
111
|
-
| ROCm Version <br/> (Variant) | vLLM
|
|
112
|
-
|
|
113
|
-
| 7.0 | `0.
|
|
114
|
-
| 6.4 | `0.
|
|
111
|
+
| ROCm Version <br/> (Variant) | vLLM | SGLang |
|
|
112
|
+
|------------------------------|-----------------------------------------------------------|-----------------------------------------------------|
|
|
113
|
+
| 7.0 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2` |
|
|
114
|
+
| 6.4 | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, `0.10.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`, `0.5.5.post3` |
|
|
115
115
|
|
|
116
116
|
## Directory Structure
|
|
117
117
|
|
|
@@ -27,8 +27,8 @@ version_tuple: VERSION_TUPLE
|
|
|
27
27
|
__commit_id__: COMMIT_ID
|
|
28
28
|
commit_id: COMMIT_ID
|
|
29
29
|
|
|
30
|
-
__version__ = version = '0.1.24.
|
|
31
|
-
__version_tuple__ = version_tuple = (0, 1, 24, '
|
|
30
|
+
__version__ = version = '0.1.24.post4'
|
|
31
|
+
__version_tuple__ = version_tuple = (0, 1, 24, 'post4')
|
|
32
32
|
try:
|
|
33
33
|
from ._version_appendix import git_commit
|
|
34
34
|
__commit_id__ = commit_id = git_commit
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
git_commit = "907c39c"
|
{gpustack_runner-0.1.24.post2 → gpustack_runner-0.1.24.post4}/gpustack_runner/runner.py.json
RENAMED
|
@@ -261,7 +261,7 @@
|
|
|
261
261
|
"service_version": "0.10.1.1",
|
|
262
262
|
"platform": "linux/amd64",
|
|
263
263
|
"docker_image": "gpustack/runner:cann8.2-a3-vllm0.10.1.1",
|
|
264
|
-
"deprecated":
|
|
264
|
+
"deprecated": true
|
|
265
265
|
},
|
|
266
266
|
{
|
|
267
267
|
"backend": "cann",
|
|
@@ -272,7 +272,7 @@
|
|
|
272
272
|
"service_version": "0.10.1.1",
|
|
273
273
|
"platform": "linux/arm64",
|
|
274
274
|
"docker_image": "gpustack/runner:cann8.2-a3-vllm0.10.1.1",
|
|
275
|
-
"deprecated":
|
|
275
|
+
"deprecated": true
|
|
276
276
|
},
|
|
277
277
|
{
|
|
278
278
|
"backend": "cann",
|
|
@@ -558,7 +558,7 @@
|
|
|
558
558
|
"service_version": "0.10.1.1",
|
|
559
559
|
"platform": "linux/amd64",
|
|
560
560
|
"docker_image": "gpustack/runner:cann8.2-910b-vllm0.10.1.1",
|
|
561
|
-
"deprecated":
|
|
561
|
+
"deprecated": true
|
|
562
562
|
},
|
|
563
563
|
{
|
|
564
564
|
"backend": "cann",
|
|
@@ -569,7 +569,7 @@
|
|
|
569
569
|
"service_version": "0.10.1.1",
|
|
570
570
|
"platform": "linux/arm64",
|
|
571
571
|
"docker_image": "gpustack/runner:cann8.2-910b-vllm0.10.1.1",
|
|
572
|
-
"deprecated":
|
|
572
|
+
"deprecated": true
|
|
573
573
|
},
|
|
574
574
|
{
|
|
575
575
|
"backend": "cann",
|
|
@@ -624,7 +624,7 @@
|
|
|
624
624
|
"service_version": "0.9.1",
|
|
625
625
|
"platform": "linux/amd64",
|
|
626
626
|
"docker_image": "gpustack/runner:cann8.2-910b-vllm0.9.1",
|
|
627
|
-
"deprecated":
|
|
627
|
+
"deprecated": true
|
|
628
628
|
},
|
|
629
629
|
{
|
|
630
630
|
"backend": "cann",
|
|
@@ -635,7 +635,7 @@
|
|
|
635
635
|
"service_version": "0.9.1",
|
|
636
636
|
"platform": "linux/arm64",
|
|
637
637
|
"docker_image": "gpustack/runner:cann8.2-910b-vllm0.9.1",
|
|
638
|
-
"deprecated":
|
|
638
|
+
"deprecated": true
|
|
639
639
|
},
|
|
640
640
|
{
|
|
641
641
|
"backend": "cann",
|
|
@@ -802,6 +802,28 @@
|
|
|
802
802
|
"docker_image": "gpustack/runner:corex4.2-vllm0.8.3",
|
|
803
803
|
"deprecated": false
|
|
804
804
|
},
|
|
805
|
+
{
|
|
806
|
+
"backend": "cuda",
|
|
807
|
+
"backend_version": "12.9",
|
|
808
|
+
"original_backend_version": "12.9.1",
|
|
809
|
+
"backend_variant": "",
|
|
810
|
+
"service": "sglang",
|
|
811
|
+
"service_version": "0.5.8",
|
|
812
|
+
"platform": "linux/amd64",
|
|
813
|
+
"docker_image": "gpustack/runner:cuda12.9-sglang0.5.8",
|
|
814
|
+
"deprecated": false
|
|
815
|
+
},
|
|
816
|
+
{
|
|
817
|
+
"backend": "cuda",
|
|
818
|
+
"backend_version": "12.9",
|
|
819
|
+
"original_backend_version": "12.9.1",
|
|
820
|
+
"backend_variant": "",
|
|
821
|
+
"service": "sglang",
|
|
822
|
+
"service_version": "0.5.8",
|
|
823
|
+
"platform": "linux/arm64",
|
|
824
|
+
"docker_image": "gpustack/runner:cuda12.9-sglang0.5.8",
|
|
825
|
+
"deprecated": false
|
|
826
|
+
},
|
|
805
827
|
{
|
|
806
828
|
"backend": "cuda",
|
|
807
829
|
"backend_version": "12.9",
|
|
@@ -846,6 +868,28 @@
|
|
|
846
868
|
"docker_image": "gpustack/runner:cuda12.9-sglang0.5.6.post2",
|
|
847
869
|
"deprecated": false
|
|
848
870
|
},
|
|
871
|
+
{
|
|
872
|
+
"backend": "cuda",
|
|
873
|
+
"backend_version": "12.9",
|
|
874
|
+
"original_backend_version": "12.9.1",
|
|
875
|
+
"backend_variant": "",
|
|
876
|
+
"service": "vllm",
|
|
877
|
+
"service_version": "0.14.1",
|
|
878
|
+
"platform": "linux/amd64",
|
|
879
|
+
"docker_image": "gpustack/runner:cuda12.9-vllm0.14.1",
|
|
880
|
+
"deprecated": false
|
|
881
|
+
},
|
|
882
|
+
{
|
|
883
|
+
"backend": "cuda",
|
|
884
|
+
"backend_version": "12.9",
|
|
885
|
+
"original_backend_version": "12.9.1",
|
|
886
|
+
"backend_variant": "",
|
|
887
|
+
"service": "vllm",
|
|
888
|
+
"service_version": "0.14.1",
|
|
889
|
+
"platform": "linux/arm64",
|
|
890
|
+
"docker_image": "gpustack/runner:cuda12.9-vllm0.14.1",
|
|
891
|
+
"deprecated": false
|
|
892
|
+
},
|
|
849
893
|
{
|
|
850
894
|
"backend": "cuda",
|
|
851
895
|
"backend_version": "12.9",
|
|
@@ -912,6 +956,28 @@
|
|
|
912
956
|
"docker_image": "gpustack/runner:cuda12.9-vllm0.11.2",
|
|
913
957
|
"deprecated": false
|
|
914
958
|
},
|
|
959
|
+
{
|
|
960
|
+
"backend": "cuda",
|
|
961
|
+
"backend_version": "12.8",
|
|
962
|
+
"original_backend_version": "12.8.1",
|
|
963
|
+
"backend_variant": "",
|
|
964
|
+
"service": "sglang",
|
|
965
|
+
"service_version": "0.5.8",
|
|
966
|
+
"platform": "linux/amd64",
|
|
967
|
+
"docker_image": "gpustack/runner:cuda12.8-sglang0.5.8",
|
|
968
|
+
"deprecated": false
|
|
969
|
+
},
|
|
970
|
+
{
|
|
971
|
+
"backend": "cuda",
|
|
972
|
+
"backend_version": "12.8",
|
|
973
|
+
"original_backend_version": "12.8.1",
|
|
974
|
+
"backend_variant": "",
|
|
975
|
+
"service": "sglang",
|
|
976
|
+
"service_version": "0.5.8",
|
|
977
|
+
"platform": "linux/arm64",
|
|
978
|
+
"docker_image": "gpustack/runner:cuda12.8-sglang0.5.8",
|
|
979
|
+
"deprecated": false
|
|
980
|
+
},
|
|
915
981
|
{
|
|
916
982
|
"backend": "cuda",
|
|
917
983
|
"backend_version": "12.8",
|
|
@@ -998,7 +1064,7 @@
|
|
|
998
1064
|
"service_version": "0.5.4.post3",
|
|
999
1065
|
"platform": "linux/amd64",
|
|
1000
1066
|
"docker_image": "gpustack/runner:cuda12.8-sglang0.5.4.post3",
|
|
1001
|
-
"deprecated":
|
|
1067
|
+
"deprecated": true
|
|
1002
1068
|
},
|
|
1003
1069
|
{
|
|
1004
1070
|
"backend": "cuda",
|
|
@@ -1009,6 +1075,28 @@
|
|
|
1009
1075
|
"service_version": "0.5.4.post3",
|
|
1010
1076
|
"platform": "linux/arm64",
|
|
1011
1077
|
"docker_image": "gpustack/runner:cuda12.8-sglang0.5.4.post3",
|
|
1078
|
+
"deprecated": true
|
|
1079
|
+
},
|
|
1080
|
+
{
|
|
1081
|
+
"backend": "cuda",
|
|
1082
|
+
"backend_version": "12.8",
|
|
1083
|
+
"original_backend_version": "12.8.1",
|
|
1084
|
+
"backend_variant": "",
|
|
1085
|
+
"service": "vllm",
|
|
1086
|
+
"service_version": "0.14.1",
|
|
1087
|
+
"platform": "linux/amd64",
|
|
1088
|
+
"docker_image": "gpustack/runner:cuda12.8-vllm0.14.1",
|
|
1089
|
+
"deprecated": false
|
|
1090
|
+
},
|
|
1091
|
+
{
|
|
1092
|
+
"backend": "cuda",
|
|
1093
|
+
"backend_version": "12.8",
|
|
1094
|
+
"original_backend_version": "12.8.1",
|
|
1095
|
+
"backend_variant": "",
|
|
1096
|
+
"service": "vllm",
|
|
1097
|
+
"service_version": "0.14.1",
|
|
1098
|
+
"platform": "linux/arm64",
|
|
1099
|
+
"docker_image": "gpustack/runner:cuda12.8-vllm0.14.1",
|
|
1012
1100
|
"deprecated": false
|
|
1013
1101
|
},
|
|
1014
1102
|
{
|
|
@@ -1209,6 +1297,28 @@
|
|
|
1209
1297
|
"docker_image": "gpustack/runner:cuda12.8-voxbox0.0.20",
|
|
1210
1298
|
"deprecated": true
|
|
1211
1299
|
},
|
|
1300
|
+
{
|
|
1301
|
+
"backend": "cuda",
|
|
1302
|
+
"backend_version": "12.6",
|
|
1303
|
+
"original_backend_version": "12.6.3",
|
|
1304
|
+
"backend_variant": "",
|
|
1305
|
+
"service": "vllm",
|
|
1306
|
+
"service_version": "0.14.1",
|
|
1307
|
+
"platform": "linux/amd64",
|
|
1308
|
+
"docker_image": "gpustack/runner:cuda12.6-vllm0.14.1",
|
|
1309
|
+
"deprecated": false
|
|
1310
|
+
},
|
|
1311
|
+
{
|
|
1312
|
+
"backend": "cuda",
|
|
1313
|
+
"backend_version": "12.6",
|
|
1314
|
+
"original_backend_version": "12.6.3",
|
|
1315
|
+
"backend_variant": "",
|
|
1316
|
+
"service": "vllm",
|
|
1317
|
+
"service_version": "0.14.1",
|
|
1318
|
+
"platform": "linux/arm64",
|
|
1319
|
+
"docker_image": "gpustack/runner:cuda12.6-vllm0.14.1",
|
|
1320
|
+
"deprecated": false
|
|
1321
|
+
},
|
|
1212
1322
|
{
|
|
1213
1323
|
"backend": "cuda",
|
|
1214
1324
|
"backend_version": "12.6",
|
|
@@ -1638,6 +1748,17 @@
|
|
|
1638
1748
|
"docker_image": "gpustack/runner:musa4.1-vllm0.9.2",
|
|
1639
1749
|
"deprecated": false
|
|
1640
1750
|
},
|
|
1751
|
+
{
|
|
1752
|
+
"backend": "rocm",
|
|
1753
|
+
"backend_version": "7.0",
|
|
1754
|
+
"original_backend_version": "7.0.2",
|
|
1755
|
+
"backend_variant": "",
|
|
1756
|
+
"service": "sglang",
|
|
1757
|
+
"service_version": "0.5.8",
|
|
1758
|
+
"platform": "linux/amd64",
|
|
1759
|
+
"docker_image": "gpustack/runner:rocm7.0-sglang0.5.8",
|
|
1760
|
+
"deprecated": false
|
|
1761
|
+
},
|
|
1641
1762
|
{
|
|
1642
1763
|
"backend": "rocm",
|
|
1643
1764
|
"backend_version": "7.0",
|
|
@@ -1660,6 +1781,17 @@
|
|
|
1660
1781
|
"docker_image": "gpustack/runner:rocm7.0-sglang0.5.6.post2",
|
|
1661
1782
|
"deprecated": false
|
|
1662
1783
|
},
|
|
1784
|
+
{
|
|
1785
|
+
"backend": "rocm",
|
|
1786
|
+
"backend_version": "7.0",
|
|
1787
|
+
"original_backend_version": "7.0.2",
|
|
1788
|
+
"backend_variant": "",
|
|
1789
|
+
"service": "vllm",
|
|
1790
|
+
"service_version": "0.14.1",
|
|
1791
|
+
"platform": "linux/amd64",
|
|
1792
|
+
"docker_image": "gpustack/runner:rocm7.0-vllm0.14.1",
|
|
1793
|
+
"deprecated": false
|
|
1794
|
+
},
|
|
1663
1795
|
{
|
|
1664
1796
|
"backend": "rocm",
|
|
1665
1797
|
"backend_version": "7.0",
|
|
@@ -1704,6 +1836,17 @@
|
|
|
1704
1836
|
"docker_image": "gpustack/runner:rocm7.0-vllm0.11.0",
|
|
1705
1837
|
"deprecated": true
|
|
1706
1838
|
},
|
|
1839
|
+
{
|
|
1840
|
+
"backend": "rocm",
|
|
1841
|
+
"backend_version": "6.4",
|
|
1842
|
+
"original_backend_version": "6.4.4",
|
|
1843
|
+
"backend_variant": "",
|
|
1844
|
+
"service": "sglang",
|
|
1845
|
+
"service_version": "0.5.8",
|
|
1846
|
+
"platform": "linux/amd64",
|
|
1847
|
+
"docker_image": "gpustack/runner:rocm6.4-sglang0.5.8",
|
|
1848
|
+
"deprecated": false
|
|
1849
|
+
},
|
|
1707
1850
|
{
|
|
1708
1851
|
"backend": "rocm",
|
|
1709
1852
|
"backend_version": "6.4",
|
|
@@ -1737,6 +1880,17 @@
|
|
|
1737
1880
|
"docker_image": "gpustack/runner:rocm6.4-sglang0.5.5.post3",
|
|
1738
1881
|
"deprecated": false
|
|
1739
1882
|
},
|
|
1883
|
+
{
|
|
1884
|
+
"backend": "rocm",
|
|
1885
|
+
"backend_version": "6.4",
|
|
1886
|
+
"original_backend_version": "6.4.4",
|
|
1887
|
+
"backend_variant": "",
|
|
1888
|
+
"service": "vllm",
|
|
1889
|
+
"service_version": "0.14.1",
|
|
1890
|
+
"platform": "linux/amd64",
|
|
1891
|
+
"docker_image": "gpustack/runner:rocm6.4-vllm0.14.1",
|
|
1892
|
+
"deprecated": false
|
|
1893
|
+
},
|
|
1740
1894
|
{
|
|
1741
1895
|
"backend": "rocm",
|
|
1742
1896
|
"backend_version": "6.4",
|
gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_sglang_reinstall_kernel/cann/Dockerfile
ADDED
|
@@ -0,0 +1,74 @@
|
|
|
1
|
+
ARG CMAKE_MAX_JOBS
|
|
2
|
+
ARG CANN_VERSION=8.3
|
|
3
|
+
ARG CANN_ARCHS=910b
|
|
4
|
+
ARG SGLANG_VERSION=0.12.0
|
|
5
|
+
ARG SGLANG_KERNEL_VERSION=20251206
|
|
6
|
+
|
|
7
|
+
FROM gpustack/runner:cann${CANN_VERSION}-${CANN_ARCHS}-sglang${SGLANG_VERSION} AS sglang
|
|
8
|
+
SHELL ["/bin/bash", "-eo", "pipefail", "-c"]
|
|
9
|
+
|
|
10
|
+
ARG TARGETPLATFORM
|
|
11
|
+
ARG TARGETOS
|
|
12
|
+
ARG TARGETARCH
|
|
13
|
+
|
|
14
|
+
## Reinstall SGLang Kernel
|
|
15
|
+
|
|
16
|
+
ARG CMAKE_MAX_JOBS
|
|
17
|
+
ARG SGLANG_VERSION
|
|
18
|
+
ARG SGLANG_KERNEL_VERSION
|
|
19
|
+
|
|
20
|
+
ENV SGLANG_VERSION=${SGLANG_VERSION} \
|
|
21
|
+
SGLANG_KERNEL_VERSION=${SGLANG_KERNEL_VERSION}
|
|
22
|
+
|
|
23
|
+
RUN <<EOF
|
|
24
|
+
# SGLang
|
|
25
|
+
|
|
26
|
+
CMAKE_MAX_JOBS="${CMAKE_MAX_JOBS}"
|
|
27
|
+
if [[ -z "${CMAKE_MAX_JOBS}" ]]; then
|
|
28
|
+
CMAKE_MAX_JOBS="$(( $(nproc) / 2 ))"
|
|
29
|
+
fi
|
|
30
|
+
if (( $(echo "${CMAKE_MAX_JOBS} > 8" | bc -l) )); then
|
|
31
|
+
CMAKE_MAX_JOBS="8"
|
|
32
|
+
fi
|
|
33
|
+
export MAX_JOBS="${CMAKE_MAX_JOBS}"
|
|
34
|
+
export COMPILE_CUSTOM_KERNELS=1
|
|
35
|
+
export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:${CANN_HOME}/ascend-toolkit/latest/$(uname -i)-linux/devlib"
|
|
36
|
+
export LD_LIBRARY_PATH="${CANN_HOME}/ascend-toolkit/latest/runtime/lib64/stub:${LD_LIBRARY_PATH}"
|
|
37
|
+
source ${CANN_HOME}/ascend-toolkit/set_env.sh
|
|
38
|
+
echo "Building SGLang with the following environment variables:"
|
|
39
|
+
env
|
|
40
|
+
|
|
41
|
+
# Install Dependencies
|
|
42
|
+
cat <<EOT >/tmp/requirements.txt
|
|
43
|
+
attrs==25.4.0
|
|
44
|
+
decorator==5.2.1
|
|
45
|
+
psutil==7.1.3
|
|
46
|
+
pyyaml==6.0.3
|
|
47
|
+
triton-ascend==3.2.0
|
|
48
|
+
EOT
|
|
49
|
+
uv pip install \
|
|
50
|
+
-r /tmp/requirements.txt
|
|
51
|
+
|
|
52
|
+
# Build and Install SGLang Kernel
|
|
53
|
+
git -C /tmp clone --recursive --shallow-submodules \
|
|
54
|
+
--depth 1 --branch ${SGLANG_KERNEL_VERSION} --single-branch \
|
|
55
|
+
https://github.com/sgl-project/sgl-kernel-npu.git sgl-kernel-npu
|
|
56
|
+
unset ASCEND_HOME_PATH
|
|
57
|
+
pushd /tmp/sgl-kernel-npu \
|
|
58
|
+
&& ./build.sh \
|
|
59
|
+
&& tree -hs /tmp/sgl-kernel-npu/output \
|
|
60
|
+
&& uv pip install /tmp/sgl-kernel-npu/output/deep_ep*.whl /tmp/sgl-kernel-npu/output/sgl_kernel_npu*.whl
|
|
61
|
+
|
|
62
|
+
# Postprocess SGLang Kernel (DeepEP)
|
|
63
|
+
cd "$(pip show deep-ep | awk '/^Location:/ {print $2}')" && ln -sf deep_ep/deep_ep_cpp*.so
|
|
64
|
+
|
|
65
|
+
# Cleanup
|
|
66
|
+
rm -rf /var/tmp/* \
|
|
67
|
+
&& rm -rf /tmp/* \
|
|
68
|
+
&& ccache --clear --clean
|
|
69
|
+
EOF
|
|
70
|
+
|
|
71
|
+
## Entrypoint
|
|
72
|
+
|
|
73
|
+
WORKDIR /
|
|
74
|
+
ENTRYPOINT [ "tini", "--" ]
|
gpustack_runner-0.1.24.post4/pack/.post_operation/20260129_sglang_reinstall_kernel/matrix.yaml
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
1
|
+
rules:
|
|
2
|
+
|
|
3
|
+
#
|
|
4
|
+
# Ascend CANN
|
|
5
|
+
#
|
|
6
|
+
|
|
7
|
+
## Packed Ascend CANN 8.3, using CANN Kernel for A3.
|
|
8
|
+
##
|
|
9
|
+
- backend: "cann"
|
|
10
|
+
services:
|
|
11
|
+
- "sglang"
|
|
12
|
+
platforms:
|
|
13
|
+
- "linux/arm64"
|
|
14
|
+
args:
|
|
15
|
+
- "CANN_VERSION=8.3"
|
|
16
|
+
- "CANN_ARCHS=a3"
|
|
17
|
+
- "SGLANG_VERSION=0.5.7"
|
|
18
|
+
## Packed Ascend CANN 8.3, using CANN Kernel for 910B.
|
|
19
|
+
##
|
|
20
|
+
- backend: "cann"
|
|
21
|
+
services:
|
|
22
|
+
- "sglang"
|
|
23
|
+
platforms:
|
|
24
|
+
- "linux/arm64"
|
|
25
|
+
args:
|
|
26
|
+
- "CANN_VERSION=8.3"
|
|
27
|
+
- "CANN_ARCHS=910b"
|
|
28
|
+
- "SGLANG_VERSION=0.5.7"
|
|
@@ -0,0 +1,25 @@
|
|
|
1
|
+
ARG CMAKE_MAX_JOBS
|
|
2
|
+
ARG CUDA_VERSION=12.8
|
|
3
|
+
ARG VLLM_VERSION=0.13.0
|
|
4
|
+
|
|
5
|
+
FROM gpustack/runner:cuda${CUDA_VERSION}-vllm${VLLM_VERSION} AS vllm
|
|
6
|
+
SHELL ["/bin/bash", "-eo", "pipefail", "-c"]
|
|
7
|
+
|
|
8
|
+
ARG TARGETPLATFORM
|
|
9
|
+
ARG TARGETOS
|
|
10
|
+
ARG TARGETARCH
|
|
11
|
+
|
|
12
|
+
## Patch
|
|
13
|
+
|
|
14
|
+
RUN --mount=type=bind,target=/workspace,rw <<EOF
|
|
15
|
+
# Patch
|
|
16
|
+
|
|
17
|
+
tree -hs /workspace/patches
|
|
18
|
+
pushd $(pip show vllm | grep Location: | cut -d" " -f 2) \
|
|
19
|
+
&& patch -p1 < /workspace/patches/vllm_*.patch
|
|
20
|
+
EOF
|
|
21
|
+
|
|
22
|
+
## Entrypoint
|
|
23
|
+
|
|
24
|
+
WORKDIR /
|
|
25
|
+
ENTRYPOINT [ "tini", "--" ]
|
|
@@ -0,0 +1,41 @@
|
|
|
1
|
+
diff --git a/vllm/utils/network_utils.py b/vllm/utils/network_utils.py
|
|
2
|
+
index 7d01533cb..311ed44df 100644
|
|
3
|
+
--- a/vllm/utils/network_utils.py
|
|
4
|
+
+++ b/vllm/utils/network_utils.py
|
|
5
|
+
@@ -147,6 +147,9 @@ def get_open_zmq_inproc_path() -> str:
|
|
6
|
+
return f"inproc://{uuid4()}"
|
|
7
|
+
|
|
8
|
+
|
|
9
|
+
+_next_port: int | None = None
|
|
10
|
+
+
|
|
11
|
+
+
|
|
12
|
+
def get_open_port() -> int:
|
|
13
|
+
"""
|
|
14
|
+
Get an open port for the vLLM process to listen on.
|
|
15
|
+
@@ -163,7 +166,7 @@ def get_open_port() -> int:
|
|
16
|
+
candidate_port = _get_open_port()
|
|
17
|
+
if candidate_port not in reserved_port_range:
|
|
18
|
+
return candidate_port
|
|
19
|
+
- return _get_open_port()
|
|
20
|
+
+ return _get_open_port(_next_port)
|
|
21
|
+
|
|
22
|
+
|
|
23
|
+
def get_open_ports_list(count: int = 5) -> list[int]:
|
|
24
|
+
@@ -174,13 +177,15 @@ def get_open_ports_list(count: int = 5) -> list[int]:
|
|
25
|
+
return list(ports)
|
|
26
|
+
|
|
27
|
+
|
|
28
|
+
-def _get_open_port() -> int:
|
|
29
|
+
- port = envs.VLLM_PORT
|
|
30
|
+
+def _get_open_port(start: int | None = None) -> int:
|
|
31
|
+
+ port = start or envs.VLLM_PORT
|
|
32
|
+
if port is not None:
|
|
33
|
+
+ global _next_port
|
|
34
|
+
while True:
|
|
35
|
+
try:
|
|
36
|
+
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
|
|
37
|
+
s.bind(("", port))
|
|
38
|
+
+ _next_port = port + 1
|
|
39
|
+
return port
|
|
40
|
+
except OSError:
|
|
41
|
+
port += 1 # Increment port number if already in use
|
|
@@ -0,0 +1,55 @@
|
|
|
1
|
+
rules:
|
|
2
|
+
|
|
3
|
+
#
|
|
4
|
+
# NVIDIA CUDA
|
|
5
|
+
#
|
|
6
|
+
|
|
7
|
+
## Packed NVIDIA CUDA 12.9.
|
|
8
|
+
##
|
|
9
|
+
- backend: "cuda"
|
|
10
|
+
services:
|
|
11
|
+
- "vllm"
|
|
12
|
+
args:
|
|
13
|
+
- "CUDA_VERSION=12.9"
|
|
14
|
+
- "VLLM_VERSION=0.13.0"
|
|
15
|
+
## Packed NVIDIA CUDA 12.8.
|
|
16
|
+
##
|
|
17
|
+
- backend: "cuda"
|
|
18
|
+
services:
|
|
19
|
+
- "vllm"
|
|
20
|
+
args:
|
|
21
|
+
- "CUDA_VERSION=12.8"
|
|
22
|
+
- "VLLM_VERSION=0.13.0"
|
|
23
|
+
## Packed NVIDIA CUDA 12.6.
|
|
24
|
+
##
|
|
25
|
+
- backend: "cuda"
|
|
26
|
+
services:
|
|
27
|
+
- "vllm"
|
|
28
|
+
args:
|
|
29
|
+
- "CUDA_VERSION=12.6"
|
|
30
|
+
- "VLLM_VERSION=0.13.0"
|
|
31
|
+
|
|
32
|
+
#
|
|
33
|
+
# AMD ROCm
|
|
34
|
+
#
|
|
35
|
+
|
|
36
|
+
## Packed AMD ROCm 7.0.
|
|
37
|
+
##
|
|
38
|
+
- backend: "rocm"
|
|
39
|
+
services:
|
|
40
|
+
- "vllm"
|
|
41
|
+
platforms:
|
|
42
|
+
- "linux/amd64"
|
|
43
|
+
args:
|
|
44
|
+
- "ROCM_VERSION=7.0"
|
|
45
|
+
- "VLLM_VERSION=0.13.0"
|
|
46
|
+
## Packed AMD ROCm 6.4.
|
|
47
|
+
##
|
|
48
|
+
- backend: "rocm"
|
|
49
|
+
services:
|
|
50
|
+
- "vllm"
|
|
51
|
+
platforms:
|
|
52
|
+
- "linux/amd64"
|
|
53
|
+
args:
|
|
54
|
+
- "ROCM_VERSION=6.4"
|
|
55
|
+
- "VLLM_VERSION=0.13.0"
|