The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Fixes (#10621)
comfyanonymouse199c8cc6758
diff --git a/comfy/quant_ops.py b/comfy/quant_ops.pyindex 5af6f118e..835fc4b8d 100644--- a/comfy/quant_ops.py+++ b/comfy/quant_ops.py@@ -131,7 +131,7 @@ class QuantizedTensor(torch.Tensor):self._layout_params = layout_paramsdef __repr__(self):- layout_name = self._layout_type.__name__+ layout_name = self._layout_typeparam_str = ", ".join(f"{k}={v}" for k, v in list(self._layout_params.items())[:2])return f"QuantizedTensor(shape={self.shape}, layout={layout_name}, {param_str})"@@ -179,7 +179,7 @@ class QuantizedTensor(torch.Tensor):attr_name = f"_layout_param_{key}"layout_params[key] = inner_tensors[attr_name]- return QuantizedTensor(inner_tensors["_q_data"], layout_type, layout_params)+ return QuantizedTensor(inner_tensors["_qdata"], layout_type, layout_params)@classmethoddef from_float(cls, tensor, layout_type, **quantize_kwargs) -> 'QuantizedTensor':