Unverified Commit a220c14f authored by scut-cbq's avatar scut-cbq Committed by GitHub
Browse files

fix crash of DeepSeek-V3 update_weights_from_disk (#8863)


Co-authored-by: default avatarparkeychen <parkeychen@tencent.com>
parent 35ef3f29
...@@ -358,8 +358,8 @@ class Fp8LinearMethod(LinearMethodBase): ...@@ -358,8 +358,8 @@ class Fp8LinearMethod(LinearMethodBase):
return return
else: else:
weight, weight_scale = layer.weight.data, layer.weight_scale_inv.data weight, weight_scale = layer.weight.data, layer.weight_scale_inv.data
layer.weight = Parameter(weight, requires_grad=False) layer.weight.data = weight.data
layer.weight_scale_inv = Parameter(weight_scale, requires_grad=False) layer.weight_scale_inv.data = weight_scale.data
else: else:
layer.weight = Parameter(layer.weight.data, requires_grad=False) layer.weight = Parameter(layer.weight.data, requires_grad=False)
......
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