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 14, 05:30 UTC
Upgrade FlashInfer to v0.3.0 (#24086)
Signed-off-by: Po-Han Huang <pohanh@nvidia.com> Co-authored-by: Simon Mo <simon.mo@hey.com>
Po-Han Huang (NVIDIA) committed
Sep 5, 2025 at 00:49 UTC
78336a0c3ee4eb9dba6e37959d926160e91623fd
2 files changed
+2
-2
docker/Dockerfile
+1
-1
@@ -375,7 +375,7 @@ RUN --mount=type=bind,from=build,src=/workspace/dist,target=/vllm-workspace/dist
375
# Install FlashInfer from source
376
ARG FLASHINFER_GIT_REPO="https://github.com/flashinfer-ai/flashinfer.git"
377
# Keep this in sync with "flashinfer" extra in setup.py
378
-ARG FLASHINFER_GIT_REF="v0.2.14.post1"
378
+ARG FLASHINFER_GIT_REF="v0.3.0"
379
# Flag to control whether to compile FlashInfer AOT kernels
380
# Set to "true" to enable AOT compilation:
381
# docker build --build-arg FLASHINFER_AOT_COMPILE=true ...
setup.py
+1
-1
@@ -694,7 +694,7 @@ setup(
694
"mistral_common[audio]"], # Required for audio processing
695
"video": [], # Kept for backwards compatibility
696
# FlashInfer should be updated together with the Dockerfile
697
- "flashinfer": ["flashinfer-python==0.2.14.post1"],
697
+ "flashinfer": ["flashinfer-python==0.3.0"],
698
# Optional deps for AMD FP4 quantization support
699
"petit-kernel": ["petit-kernel"],
700
},