[Build][NVIDIA] Update public Rubin dependencies and MSA compatibility (#56545)
Signed-off-by: Shang Wang <samshang.wang@mail.utoronto.ca> Co-authored-by: Codex <noreply@openai.com>
This commit is contained in:
@@ -17,7 +17,7 @@ else()
|
|||||||
FetchContent_Declare(
|
FetchContent_Declare(
|
||||||
fmha_sm100
|
fmha_sm100
|
||||||
GIT_REPOSITORY https://github.com/vllm-project/MSA.git
|
GIT_REPOSITORY https://github.com/vllm-project/MSA.git
|
||||||
GIT_TAG 087c161814d4d9c735b46c21212a09e5f8eb92fa
|
GIT_TAG f355c37eb4e1413f21ee2ad8bbad25079e6bef9d
|
||||||
GIT_PROGRESS TRUE
|
GIT_PROGRESS TRUE
|
||||||
CONFIGURE_COMMAND ""
|
CONFIGURE_COMMAND ""
|
||||||
BUILD_COMMAND ""
|
BUILD_COMMAND ""
|
||||||
|
|||||||
+1
-1
@@ -641,7 +641,7 @@ RUN --mount=type=cache,target=/opt/uv/cache \
|
|||||||
python3 use_existing_torch.py --prefix; \
|
python3 use_existing_torch.py --prefix; \
|
||||||
fi \
|
fi \
|
||||||
&& if [ "${INSTALL_RUBIN_PRERELEASE}" = "true" ]; then \
|
&& if [ "${INSTALL_RUBIN_PRERELEASE}" = "true" ]; then \
|
||||||
sed -Ei '/^[[:space:]]*(flashinfer-python|nvidia-cutlass-dsl\[cu13\])[[:space:]]*==/Id' \
|
sed -Ei '/^[[:space:]]*(flashinfer-python|nvidia-cutlass-dsl\[cu13\]|quack-kernels)[[:space:]]*==/Id' \
|
||||||
requirements/cuda.txt; \
|
requirements/cuda.txt; \
|
||||||
fi
|
fi
|
||||||
|
|
||||||
|
|||||||
@@ -2,15 +2,25 @@
|
|||||||
# Installed when INSTALL_RUBIN_PRERELEASE=true (see docker/Dockerfile).
|
# Installed when INSTALL_RUBIN_PRERELEASE=true (see docker/Dockerfile).
|
||||||
# TODO(rubin-bringup): delete once Rubin APIs ship in regular PyTorch wheels
|
# TODO(rubin-bringup): delete once Rubin APIs ship in regular PyTorch wheels
|
||||||
# (vllm-project/vllm#49735).
|
# (vllm-project/vllm#49735).
|
||||||
--pre
|
|
||||||
--extra-index-url https://download.pytorch.org/whl/nightly/cu134
|
--extra-index-url https://download.pytorch.org/whl/nightly/cu134
|
||||||
|
--extra-index-url https://flashinfer.ai/whl/cu134
|
||||||
|
--extra-index-url https://flashinfer.ai/whl
|
||||||
|
|
||||||
torch==2.15.0.dev20260903+cu134
|
# Match each architecture's torch pin to the latest torchvision wheel.
|
||||||
torchvision==0.30.0.dev20260903+cu134
|
torch==2.15.0.dev20260909+cu134; platform_machine == "x86_64"
|
||||||
torchaudio==2.11.0.dev20260903+cu134
|
torch==2.15.0.dev20260908+cu134; platform_machine == "aarch64"
|
||||||
|
torchvision==0.30.0.dev20260909+cu134
|
||||||
|
torchaudio==2.11.0.dev20260911+cu134
|
||||||
|
|
||||||
# docker/Dockerfile strips these two from requirements/cuda.txt whenever
|
# docker/Dockerfile strips these packages from requirements/cuda.txt whenever
|
||||||
# INSTALL_RUBIN_PRERELEASE=true, so they must be pinned here instead to
|
# INSTALL_RUBIN_PRERELEASE=true, so they must be pinned here instead to
|
||||||
# still get installed.
|
# still get installed.
|
||||||
nvidia-cutlass-dsl[cu13]==4.8.0.dev0
|
nvidia-cutlass-dsl[cu13]==4.8.0.dev0
|
||||||
flashinfer-python==0.6.18
|
flashinfer-python==0.6.18.post1
|
||||||
|
quack-kernels==0.6.5
|
||||||
|
|
||||||
|
cuda-python==13.4.1
|
||||||
|
cuda-bindings==13.4.1
|
||||||
|
flashinfer-jit-cache==0.6.18.post1+cu134
|
||||||
|
flashinfer-cubin==0.6.18.post1
|
||||||
|
nixl==1.4.1
|
||||||
|
|||||||
Reference in New Issue
Block a user