subject fill

README.md

@@ -4,7 +4,7 @@ emoji: 🌍
 colorFrom: blue
 colorTo: green
 sdk: gradio
-sdk_version: 5.23.2
+sdk_version: 5.6.0
 app_file: app.py
 pinned: false
 ---

app.py

@@ -97,65 +97,31 @@ header = """
 
 
 def create_app():
-    # with gr.Blocks() as app:
-    #     gr.Markdown(header, elem_id="header")
-    #     # with gr.Tabs():
-    #     #     with gr.Tab("Subject-driven"):
-    #     gr.Interface(
-    #         fn=process_image_and_text,
-    #         inputs=[
-    #             gr.Image(type="pil", label="Condition Image", width=300, elem_id="input"),
-    #             gr.Radio(
-    #                 [("512", 512), ("1024(beta)", 1024)],
-    #                 label="Resolution",
-    #                 value=512,
-    #                 elem_id="resolution",
-    #             ),
-    #             # gr.Slider(4, 16, 4, step=4, label="Inference Steps"),
-    #             gr.Textbox(lines=2, label="Text Prompt", elem_id="text"),
-    #         ],
-    #         outputs=gr.Image(type="pil", elem_id="output"),
-    #         examples=get_samples(),
-    #     )
-    #         # with gr.Tab("Fill"):
-    #         #     gr.Markdown("Coming soon")
-    #         # with gr.Tab("Canny"):
-    #         #     gr.Markdown("Coming soon")
-    #         # with gr.Tab("Depth"):
-    #         #     gr.Markdown("Coming soon")
-    
     with gr.Blocks() as app:
-        gr.Markdown(header, elem_id="header")
-        with gr.Row(equal_height=False):
-            with gr.Column(variant="panel", elem_classes="inputPanel"):
-                original_image = gr.Image(
-                    type="pil", label="Condition Image", width=300, elem_id="input"
+        gr.Markdown(header)
+        with gr.Tabs():
+            with gr.Tab("Subject-driven"):
+                gr.Interface(
+                    fn=process_image_and_text,
+                    inputs=[
+                        gr.Image(type="pil", label="Condition Image", width=300),
+                        gr.Radio(
+                            [("512", 512), ("1024(beta)", 1024)],
+                            label="Resolution",
+                            value=512,
+                        ),
+                        # gr.Slider(4, 16, 4, step=4, label="Inference Steps"),
+                        gr.Textbox(lines=2, label="Text Prompt"),
+                    ],
+                    outputs=gr.Image(type="pil"),
+                    examples=get_samples(),
                 )
-                resolution = gr.Radio(
-                    [("512", 512), ("1024(beta)", 1024)],
-                    label="Resolution",
-                    value=512,
-                    elem_id="resolution",
-                )
-                text = gr.Textbox(lines=2, label="Text Prompt", elem_id="text")
-                submit_btn = gr.Button("Run", elem_id="submit_btn")
-
-            with gr.Column(variant="panel", elem_classes="outputPanel"):
-                output_image = gr.Image(type="pil", elem_id="output")
-
-        with gr.Row():
-            examples = gr.Examples(
-                examples=get_samples(),
-                inputs=[original_image, resolution, text],
-                label="Examples",
-            )
-
-        submit_btn.click(
-            fn=process_image_and_text,
-            inputs=[original_image, resolution, text],
-            outputs=output_image,
-        )
-        
+            with gr.Tab("Fill"):
+                gr.Markdown("Coming soon")
+            with gr.Tab("Canny"):
+                gr.Markdown("Coming soon")
+            with gr.Tab("Depth"):
+                gr.Markdown("Coming soon")
     return app
 
 

src/transformer.py

@@ -7,6 +7,7 @@ from diffusers.models.transformers.transformer_flux import (
     FluxTransformer2DModel,
     Transformer2DModelOutput,
     USE_PEFT_BACKEND,
+    is_torch_version,
     scale_lora_layers,
     unscale_lora_layers,
     logger,
@@ -154,6 +155,9 @@ def tranformer_forward(
 
                 return custom_forward
 
+            ckpt_kwargs: Dict[str, Any] = (
+                {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
+            )
             encoder_hidden_states, hidden_states = torch.utils.checkpoint.checkpoint(
                 create_custom_forward(block),
                 hidden_states,
@@ -200,6 +204,9 @@ def tranformer_forward(
 
                 return custom_forward
 
+            ckpt_kwargs: Dict[str, Any] = (
+                {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
+            )
             hidden_states = torch.utils.checkpoint.checkpoint(
                 create_custom_forward(block),
                 hidden_states,