xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-06 03:17:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

add minimal documentation for torch scaled mm base class

Browse Source 
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
This commit is contained in:
vllmellm 2025-11-07 07:30:57 +00:00
parent 7fb465744c
commit 56a05cd818
1 changed files with 8 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/layers/quantization/kernels/scaled_mm/pytorch.py
View File

@ -133,6 +133,14 @@ def torch_channelwise_w8a8_scaled_mm(
class TorchScaledMMLinearKernel(FP8ScaledMMLinearKernel):
"""
Base class for FP8 linear kernels using Torch.
Each subclass represents a kernel variant for
specific device capabilities and torch versions,
so we split them up and implement
get_min_capability() separately for each.
"""
def get_ouput_padding(self) -> int | None:
vllm_config = get_current_vllm_config().compilation_config
pad_output = vllm_config.mode < CompilationMode.VLLM_COMPILE