Skip to content

GitLab

Unverified Commit 5be4fe5a authored by Kane's avatar Kane Committed by GitHub
Browse files

fix mlu int8 quant (#531) 

1. 修复mlu int8量化
parent f7665abb
Hide whitespace changes
Inline Side-by-side
Showing with 3 additions and 1 deletion
lightx2v/common/ops/mm/mm_weight.py
View file @ 5be4fe5a
...@@ -1204,5 +1204,7 @@ class MMWeightWint8channelAint8channeldynamicMlu(MMWeightQuantTemplate): ...@@ -1204,5 +1204,7 @@ class MMWeightWint8channelAint8channeldynamicMlu(MMWeightQuantTemplate):
def apply(self, input_tensor): def apply(self, input_tensor):
dtype = input_tensor.dtype dtype = input_tensor.dtype
input_tensor_quant, input_tensor_scale = self.act_quant_func(input_tensor) input_tensor_quant, input_tensor_scale = self.act_quant_func(input_tensor)
output_tensor = tmo.scaled_matmul(input_tensor_quant, self.weight.contiguous(), input_tensor_scale, self.weight_scale.squeeze(-1), output_dtype=dtype, use_hp_active=True) output_tensor = tmo.scaled_matmul(
input_tensor_quant, self.weight.contiguous(), input_tensor_scale, self.weight_scale.squeeze(-1), bias=self.bias if self.bias is not None else None, output_dtype=dtype, use_hp_active=True
)
return output_tensor return output_tensor
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