xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-06-06 02:35:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

Revert "Fix non-contiguous input passed to Marlin kernel (#15319)" (#15398)

Browse Source
This commit is contained in:
Tyler Michael Smith 2025-03-24 23:43:51 -04:00 committed by GitHub
parent 6db94571d7
commit b5269db959
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 0 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/layers/quantization/kernels/mixed_precision/marlin.py
View File

@ -115,10 +115,6 @@ class MarlinLinearKernel(MPLinearKernel):
layer: torch.nn.Module, layer: torch.nn.Module,
x: torch.Tensor, x: torch.Tensor,
bias: Optional[torch.Tensor] = None) -> torch.Tensor: bias: Optional[torch.Tensor] = None) -> torch.Tensor:
# marlin requires contiguous memory layout
# prefix caching may cause x to be non-contiguous
x = x.contiguous() # no-op if already contiguous
c = self.config c = self.config
w_q, w_s, w_zp, w_gidx = self._get_weight_params(layer) w_q, w_s, w_zp, w_gidx = self._get_weight_params(layer)