Fix some torch compile + fp8_fast issues

2025-10-05 12:31:45 +03:00
parent 3167e088ed
commit d3010e37ab
1 changed files with 1 additions and 1 deletions
@@ -15,7 +15,7 @@ def fp8_linear_forward(cls, base_dtype, input):
            else:
                scale_weight = scale_weight.to(input.device).squeeze()
            
-            scale_input = torch.ones((), device=input.device, dtype=torch.float32).squeeze()
+            scale_input = torch.ones((), device=input.device, dtype=torch.float32)
            
            input = torch.clamp(input, min=-448, max=448, out=input)
            inn = input.reshape(-1, input_shape[2]).to(torch.float8_e4m3fn).contiguous() #always e4m3fn because e5m2 * e5m2 is not supported