diffsynth-engine 0.6.1.dev1__py3-none-any.whl → 0.6.1.dev2__py3-none-any.whl

diffsynth_engine/models/wan/wan_text_encoder.py

@@ -198,22 +198,22 @@ class WanTextEncoderStateDictConverter(StateDictConverter):
 
     def _from_diffusers(self, state_dict):
         rename_dict = {
-            "enc.output_norm.weight": "norm.weight",
-            "token_embd.weight": "token_embedding.weight",
+            "shared.weight": "token_embedding.weight",
+            "encoder.final_layer_norm.weight": "norm.weight",
         }
         for i in range(self.num_encoder_layers):
             rename_dict.update(
                 {
-                    f"enc.blk.{i}.attn_q.weight": f"blocks.{i}.attn.q.weight",
-                    f"enc.blk.{i}.attn_k.weight": f"blocks.{i}.attn.k.weight",
-                    f"enc.blk.{i}.attn_v.weight": f"blocks.{i}.attn.v.weight",
-                    f"enc.blk.{i}.attn_o.weight": f"blocks.{i}.attn.o.weight",
-                    f"enc.blk.{i}.ffn_up.weight": f"blocks.{i}.ffn.fc1.weight",
-                    f"enc.blk.{i}.ffn_down.weight": f"blocks.{i}.ffn.fc2.weight",
-                    f"enc.blk.{i}.ffn_gate.weight": f"blocks.{i}.ffn.gate.0.weight",
-                    f"enc.blk.{i}.attn_norm.weight": f"blocks.{i}.norm1.weight",
-                    f"enc.blk.{i}.ffn_norm.weight": f"blocks.{i}.norm2.weight",
-                    f"enc.blk.{i}.attn_rel_b.weight": f"blocks.{i}.pos_embedding.embedding.weight",
+                    f"encoder.block.{i}.layer.0.SelfAttention.q.weight": f"blocks.{i}.attn.q.weight",
+                    f"encoder.block.{i}.layer.0.SelfAttention.k.weight": f"blocks.{i}.attn.k.weight",
+                    f"encoder.block.{i}.layer.0.SelfAttention.v.weight": f"blocks.{i}.attn.v.weight",
+                    f"encoder.block.{i}.layer.0.SelfAttention.o.weight": f"blocks.{i}.attn.o.weight",
+                    f"encoder.block.{i}.layer.0.SelfAttention.relative_attention_bias.weight": f"blocks.{i}.pos_embedding.embedding.weight",
+                    f"encoder.block.{i}.layer.0.layer_norm.weight": f"blocks.{i}.norm1.weight",
+                    f"encoder.block.{i}.layer.1.DenseReluDense.wi_0.weight": f"blocks.{i}.ffn.gate.0.weight",
+                    f"encoder.block.{i}.layer.1.DenseReluDense.wi_1.weight": f"blocks.{i}.ffn.fc1.weight",
+                    f"encoder.block.{i}.layer.1.DenseReluDense.wo.weight": f"blocks.{i}.ffn.fc2.weight",
+                    f"encoder.block.{i}.layer.1.layer_norm.weight": f"blocks.{i}.norm2.weight",
                 }
             )
 
@@ -224,7 +224,7 @@ class WanTextEncoderStateDictConverter(StateDictConverter):
         return new_state_dict
 
     def convert(self, state_dict):
-        if "enc.output_norm.weight" in state_dict:
+        if "encoder.final_layer_norm.weight" in state_dict:
             logger.info("use diffusers format state dict")
             return self._from_diffusers(state_dict)
         return state_dict

{diffsynth_engine-0.6.1.dev1.dist-info → diffsynth_engine-0.6.1.dev2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffsynth_engine
-Version: 0.6.1.dev1
+Version: 0.6.1.dev2
 Author: MuseAI x ModelScope
 Classifier: Programming Language :: Python :: 3
 Classifier: Operating System :: OS Independent

{diffsynth_engine-0.6.1.dev1.dist-info → diffsynth_engine-0.6.1.dev2.dist-info}/RECORD

@@ -136,7 +136,7 @@ diffsynth_engine/models/wan/wan_audio_encoder.py,sha256=W7sUo3Fse33MNsZTYiYYEe0Y
 diffsynth_engine/models/wan/wan_dit.py,sha256=ocCAYl3xP0UdB9dhz9R1pRfBBESEHxDUu116hORe39U,20080
 diffsynth_engine/models/wan/wan_image_encoder.py,sha256=VE7crdTxOFN2UCMN2cQlvHB9BilSbKOBQYgnXgl4E2Y,14313
 diffsynth_engine/models/wan/wan_s2v_dit.py,sha256=sOJsSs1snI-ZGPJS8utstmgj0wcYwloyZ0C14hQHFkg,23597
-diffsynth_engine/models/wan/wan_text_encoder.py,sha256=Zv0q7Rc0Uq3SM0i1PokdmaKK5vhhJogpXxpq1tYHrtU,10768
+diffsynth_engine/models/wan/wan_text_encoder.py,sha256=OERlmwOqthAFPNnnT2sXJ4OjyyRmsRLx7VGp1zlBkLU,11021
 diffsynth_engine/models/wan/wan_vae.py,sha256=ogXrVlwmzXR4iLxjSCkBPtYW8KWebnvvd2UtPZeoziY,38853
 diffsynth_engine/pipelines/__init__.py,sha256=jh-4LSJ0vqlXiT8BgFgRIQxuAr2atEPyHrxXWj-Ud1U,604
 diffsynth_engine/pipelines/base.py,sha256=RTkVwWaWXr5ujqn5-UBHvdPddYwr-uvChj9-fmoXrms,13729
@@ -185,8 +185,8 @@ diffsynth_engine/utils/video.py,sha256=GoMyc2as4_VqfWX4pjQyAWh9QObsFMov42zADVZNa
 diffsynth_engine/utils/memory/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/utils/memory/linear_regression.py,sha256=oW_EQEw13oPoyUrxiL8A7Ksa5AuJ2ynI2qhCbfAuZbg,3930
 diffsynth_engine/utils/memory/memory_predcit_model.py,sha256=EXprSl_zlVjgfMWNXP-iw83Ot3hyMcgYaRPv-dvyL84,3943
-diffsynth_engine-0.6.1.dev1.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
-diffsynth_engine-0.6.1.dev1.dist-info/METADATA,sha256=nLek_Y6zJxZ3ycgtAGGkw8pyFfPkFapyEx1J2jmdiYI,1163
-diffsynth_engine-0.6.1.dev1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-diffsynth_engine-0.6.1.dev1.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
-diffsynth_engine-0.6.1.dev1.dist-info/RECORD,,
+diffsynth_engine-0.6.1.dev2.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
+diffsynth_engine-0.6.1.dev2.dist-info/METADATA,sha256=6bwBcpLV0q3yBPFYlDecEtAJP1ga-zi7u9tioW41D7k,1163
+diffsynth_engine-0.6.1.dev2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+diffsynth_engine-0.6.1.dev2.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
+diffsynth_engine-0.6.1.dev2.dist-info/RECORD,,