Properly save mixed ops. (#11772)

comfy/ops.py

@@ -625,21 +625,29 @@ def mixed_precision_ops(quant_config={}, compute_dtype=torch.bfloat16, full_prec
                         missing_keys.remove(key)
 
             def state_dict(self, *args, destination=None, prefix="", **kwargs):
-                sd = super().state_dict(*args, destination=destination, prefix=prefix, **kwargs)
+                if destination is not None:
-                if isinstance(self.weight, QuantizedTensor):
+                    sd = destination
-                    layout_cls = self.weight._layout_cls
+                else:
+                    sd = {}
 
-                    # Check if it's any FP8 variant (E4M3 or E5M2)
+                if self.bias is not None:
-                    if layout_cls in ("TensorCoreFP8E4M3Layout", "TensorCoreFP8E5M2Layout", "TensorCoreFP8Layout"):
+                    sd["{}bias".format(prefix)] = self.bias
-                        sd["{}weight_scale".format(prefix)] = self.weight._params.scale
+
-                    elif layout_cls == "TensorCoreNVFP4Layout":
+                if isinstance(self.weight, QuantizedTensor):
-                        sd["{}weight_scale_2".format(prefix)] = self.weight._params.scale
+                    sd_out = self.weight.state_dict("{}weight".format(prefix))
-                        sd["{}weight_scale".format(prefix)] = self.weight._params.block_scale
+                    for k in sd_out:
+                        sd[k] = sd_out[k]
 
                     quant_conf = {"format": self.quant_format}
                     if self._full_precision_mm_config:
                         quant_conf["full_precision_matrix_mult"] = True
                     sd["{}comfy_quant".format(prefix)] = torch.tensor(list(json.dumps(quant_conf).encode('utf-8')), dtype=torch.uint8)
+
+                    input_scale = getattr(self, 'input_scale', None)
+                    if input_scale is not None:
+                        sd["{}input_scale".format(prefix)] = input_scale
+                else:
+                    sd["{}weight".format(prefix)] = self.weight
                 return sd
 
             def _forward(self, input, weight, bias):

tests-unit/comfy_quant/test_mixed_precision.py

@@ -153,9 +153,9 @@ class TestMixedPrecisionOps(unittest.TestCase):
         state_dict2 = model.state_dict()
 
         # Verify layer1.weight is a QuantizedTensor with scale preserved
-        self.assertIsInstance(state_dict2["layer1.weight"], QuantizedTensor)
+        self.assertTrue(torch.equal(state_dict2["layer1.weight"].view(torch.uint8), fp8_weight.view(torch.uint8)))
-        self.assertEqual(state_dict2["layer1.weight"]._params.scale.item(), 3.0)
+        self.assertEqual(state_dict2["layer1.weight_scale"].item(), 3.0)
-        self.assertEqual(state_dict2["layer1.weight"]._layout_cls, "TensorCoreFP8E4M3Layout")
+        self.assertEqual(model.layer1.weight._layout_cls, "TensorCoreFP8E4M3Layout")
 
         # Verify non-quantized layers are standard tensors
         self.assertNotIsInstance(state_dict2["layer2.weight"], QuantizedTensor)