Skip to content

GitLab

Commit 9d44744c authored by SAC_fanth's avatar SAC_fanth
Browse files

修复channel-wise 用triton报错的bug

parent d146a231
Show whitespace changes
Inline Side-by-side
Showing with 2 additions and 3 deletions
vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py
View file @ 9d44744c
...@@ -161,9 +161,8 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme): ...@@ -161,9 +161,8 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
) )
if envs.VLLM_W8A8_BACKEND == 3: if envs.VLLM_W8A8_BACKEND == 3:
weight = weight.t().contiguous() weight = weight.t().contiguous()
# triton不用转置，torch需要 else:
# else: weight = weight.t()
# weight = weight.t()
elif self.strategy == QuantizationStrategy.BLOCK: elif self.strategy == QuantizationStrategy.BLOCK:
assert self.is_static_input_scheme is False assert self.is_static_input_scheme is False
weight, weight_scale = process_fp8_weight_block_strategy( weight, weight_scale = process_fp8_weight_block_strategy(
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment