Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Fix issue with fp8 ops on some models. (#8045)

comfyanonymousd42613686f3d
diff --git a/comfy/ops.py b/comfy/ops.py
index 032787915..431c8f89d 100644
--- a/comfy/ops.py
+++ b/comfy/ops.py
@@ -308,10 +308,10 @@ def fp8_linear(self, input):
if scale_input is None:
scale_input = torch.ones((), device=input.device, dtype=torch.float32)
input = torch.clamp(input, min=-448, max=448, out=input)
- input = input.reshape(-1, input_shape[2]).to(dtype)
+ input = input.reshape(-1, input_shape[2]).to(dtype).contiguous()
else:
scale_input = scale_input.to(input.device)
- input = (input * (1.0 / scale_input).to(input_dtype)).reshape(-1, input_shape[2]).to(dtype)
+ input = (input * (1.0 / scale_input).to(input_dtype)).reshape(-1, input_shape[2]).to(dtype).contiguous()
if bias is not None:
o = torch._scaled_mm(input, w, out_dtype=input_dtype, bias=bias, scale_a=scale_input, scale_b=scale_weight)