diffsynth-engine 0.6.1.dev17__py3-none-any.whl → 0.6.1.dev19__py3-none-any.whl

diffsynth_engine/pipelines/flux_image.py

@@ -983,8 +983,9 @@ class FluxImagePipeline(BasePipeline):
         elif self.ip_adapter is not None:
             image_emb = self.ip_adapter.encode_image(ref_image)
         elif self.redux is not None:
-            image_prompt_embeds = self.redux(ref_image)
-            positive_prompt_emb = torch.cat([positive_prompt_emb, image_prompt_embeds], dim=1)
+            ref_prompt_embeds = self.redux(ref_image)
+            flattened_ref_emb = ref_prompt_embeds.view(1, -1, ref_prompt_embeds.size(-1))
+            positive_prompt_emb = torch.cat([positive_prompt_emb, flattened_ref_emb], dim=1)
 
         # Extra input
         image_ids, text_ids, guidance = self.prepare_extra_input(

diffsynth_engine/pipelines/qwen_image.py

@@ -147,9 +147,10 @@ class QwenImagePipeline(BasePipeline):
         self.prompt_template_encode = "<|im_start|>system\nDescribe the image by detailing the color, shape, size, texture, quantity, text, spatial relationships of the objects and background:<|im_end|>\n<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
         self.prompt_template_encode_start_idx = 34
         # qwen image edit
-        self.edit_prompt_template_encode = "<|im_start|>system\nDescribe the key features of the input image (color, shape, size, texture, objects, background), then explain how the user's text instruction should alter or modify the image. Generate a new image that meets the user's requirements while maintaining consistency with the original input where appropriate.<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n<|im_start|>assistant\n"
+        self.edit_system_prompt = "Describe the key features of the input image (color, shape, size, texture, objects, background), then explain how the user's text instruction should alter or modify the image. Generate a new image that meets the user's requirements while maintaining consistency with the original input where appropriate."
+        self.edit_prompt_template_encode = "<|im_start|>system\n" + self.edit_system_prompt + "<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n<|im_start|>assistant\n"
         # qwen image edit plus
-        self.edit_plus_prompt_template_encode = "<|im_start|>system\nDescribe the key features of the input image (color, shape, size, texture, objects, background), then explain how the user's text instruction should alter or modify the image. Generate a new image that meets the user's requirements while maintaining consistency with the original input where appropriate.<|im_end|>\n<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
+        self.edit_plus_prompt_template_encode = "<|im_start|>system\n" + self.edit_system_prompt + "<|im_end|>\n<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
 
         self.edit_prompt_template_encode_start_idx = 64
 

{diffsynth_engine-0.6.1.dev17.dist-info → diffsynth_engine-0.6.1.dev19.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffsynth_engine
-Version: 0.6.1.dev17
+Version: 0.6.1.dev19
 Author: MuseAI x ModelScope
 Classifier: Programming Language :: Python :: 3
 Classifier: Operating System :: OS Independent

{diffsynth_engine-0.6.1.dev17.dist-info → diffsynth_engine-0.6.1.dev19.dist-info}/RECORD

@@ -140,9 +140,9 @@ diffsynth_engine/models/wan/wan_text_encoder.py,sha256=OERlmwOqthAFPNnnT2sXJ4Ojy
 diffsynth_engine/models/wan/wan_vae.py,sha256=dC7MoUFeXRL7SIY0LG1OOUiZW-pp9IbXCghutMxpXr4,38889
 diffsynth_engine/pipelines/__init__.py,sha256=jh-4LSJ0vqlXiT8BgFgRIQxuAr2atEPyHrxXWj-Ud1U,604
 diffsynth_engine/pipelines/base.py,sha256=B6Md10eeAK4itILjx3biRCFwYk2usgSv7v2V9vd4fjA,14842
-diffsynth_engine/pipelines/flux_image.py,sha256=xoCZA2IbxARjETcni3AL6B5KcNb-RqABQA1FULi5msw,50833
+diffsynth_engine/pipelines/flux_image.py,sha256=Dpy8AkwywuLAhvJ6cjg5TgzhSUgFQtv6p2JTTkzUHbo,50919
 diffsynth_engine/pipelines/hunyuan3d_shape.py,sha256=TNV0Wr09Dj2bzzlpua9WioCClOj3YiLfE6utI9aWL8A,8164
-diffsynth_engine/pipelines/qwen_image.py,sha256=o16Dffrhwel5SzCmBTE4-XnIeoq6fm3YwdPWbdsk2as,33257
+diffsynth_engine/pipelines/qwen_image.py,sha256=jt4rg-U5qWsFD0kUeDwKzgIiTAC80Cj8aq1YQOR1_-k,33052
 diffsynth_engine/pipelines/sd_image.py,sha256=nr-Nhsnomq8CsUqhTM3i2l2zG01YjwXdfRXgr_bC3F0,17891
 diffsynth_engine/pipelines/sdxl_image.py,sha256=v7ZACGPb6EcBunL6e5E9jynSQjE7GQx8etEV-ZLP91g,21704
 diffsynth_engine/pipelines/utils.py,sha256=lk7sFGEk-fGjgadLpwwppHKG-yZ0RC-4ZmHW7pRRe8A,473
@@ -185,8 +185,8 @@ diffsynth_engine/utils/video.py,sha256=8FCaeqIdUsWMgWI_6SO9SPynsToGcLCQAVYFTc4CD
 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.dev17.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
-diffsynth_engine-0.6.1.dev17.dist-info/METADATA,sha256=PypFK6NGQju_6lQq1kfFjGFsUnwft8v8MMbVjL8d_mg,1164
-diffsynth_engine-0.6.1.dev17.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-diffsynth_engine-0.6.1.dev17.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
-diffsynth_engine-0.6.1.dev17.dist-info/RECORD,,
+diffsynth_engine-0.6.1.dev19.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
+diffsynth_engine-0.6.1.dev19.dist-info/METADATA,sha256=KQ9a1ITP4r5RnWNUKEGJnnt5dduwknR3rCU2K5ETBC4,1164
+diffsynth_engine-0.6.1.dev19.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+diffsynth_engine-0.6.1.dev19.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
+diffsynth_engine-0.6.1.dev19.dist-info/RECORD,,