biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[ROCm] [Feature] [Doc] [Dockerfile] [BugFix] Support Per-Token-Activation Per-Channel-Weight FP8 Quantization Inferencing (#12501)

Browse Source
This commit is contained in:
TJian
2025-02-08 00:13:43 +08:00
committed by GitHub
parent 0630d4537a
commit eaa92d4437
8 changed files with 295 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
Dockerfile.rocm_base
View File

@@ -6,7 +6,7 @@ ARG RCCL_BRANCH="648a58d"
ARG RCCL_REPO="https://github.com/ROCm/rccl"
ARG TRITON_BRANCH="e5be006"
ARG TRITON_REPO="https://github.com/triton-lang/triton.git"
ARG PYTORCH_BRANCH="8d4926e"
ARG PYTORCH_BRANCH="3a585126"
ARG PYTORCH_VISION_BRANCH="v0.19.1"
ARG PYTORCH_REPO="https://github.com/pytorch/pytorch.git"
ARG PYTORCH_VISION_REPO="https://github.com/pytorch/vision.git"