[Bugfix] compressed-tensors: restore int8 grouped WNA16 MoE support (#52002)
Signed-off-by: Yohann Prigent <prigent.yohann@gmail.com> Co-authored-by: Misha Goin <mgoin64@gmail.com>
Yohann Prigent committed
Aug 19, 2026 at 16:55 UTC
be06873198cd544f963b4e2f0f0877b23ab5861e
1 file changed
-1
vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe/compressed_tensors_moe_wna16.py
-1
@@ -82,7 +82,6 @@ class CompressedTensorsWNA16MoEMethod(CompressedTensorsMoEMethod):
82
else:
83
scale = kInt4StaticGroupScale
84
elif self.num_bits == 8:
85
- assert self.group_size == -1
85
scale = kInt8StaticGroupScale
86
else:
87
raise ValueError(