Initial Commit

app.py

@@ -87,25 +87,9 @@ class FluxInterface:
             vae=vae,
             text_encoder=text_encoder,
             tokenizer=tokenizer,
-        ) 
-
-    def resize_image(self, img, max_pixels=1050000):
-        if not isinstance(img, Image.Image):
-            img = Image.fromarray(img)
-        
-        width, height = img.size
-        num_pixels = width * height
-        
-        if num_pixels > max_pixels:
-            scale = math.sqrt(max_pixels / num_pixels)
-            new_width = int(width * scale)
-            new_height = int(height * scale)
-            new_width = new_width - (new_width % 8)
-            new_height = new_height - (new_height % 8)
-            img = img.resize((new_width, new_height), Image.LANCZOS)
-        
-        return img
+        )
 
+    # [Previous methods remain unchanged...]
     def process_image(self, image):
         message = [
             {
@@ -125,8 +109,8 @@ class FluxInterface:
             image_hidden_state = self.models['connector'](image_hidden_state)
 
         return image_hidden_state, image_grid_thw
-    
-    def compute_t5_text_embeddings(self, prompt):
+
+    def compute_text_embeddings(self, prompt):
         """Compute T5 embeddings for text prompt"""
         if prompt == "":
             return None
@@ -145,24 +129,6 @@ class FluxInterface:
         
         return prompt_embeds
 
-    def compute_text_embeddings(self, prompt=""):
-        with torch.no_grad():
-            text_inputs = self.models['tokenizer'](
-                prompt,
-                padding="max_length",
-                max_length=77,
-                truncation=True,
-                return_tensors="pt"
-            ).to(self.device)
-
-            prompt_embeds = self.models['text_encoder'](
-                text_inputs.input_ids,
-                output_hidden_states=False
-            )
-            pooled_prompt_embeds = prompt_embeds.pooler_output.to(self.dtype)
-
-        return pooled_prompt_embeds
-
     def generate(self, input_image, prompt="", guidance_scale=3.5, num_inference_steps=28, num_images=2, seed=None):
         try:
             if seed is not None:
@@ -197,7 +163,7 @@ class FluxInterface:
 interface = FluxInterface()
 
 # Create Gradio interface
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
+with gr.Blocks() as demo:
     gr.Markdown("""
     # 🎨 Qwen2vl-Flux Image Variation Demo
     Upload an image and get AI-generated variations. You can optionally add a text prompt to guide the generation.
@@ -208,80 +174,50 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
             input_image = gr.Image(
                 label="Upload Image",
                 type="pil",
-                height=384,
-                width=384,
-                tool="select"
+                height=384
             )
             prompt = gr.Textbox(
-                label="Optional Text Prompt",
+                label="Optional Text Prompt, As Long As Possible",
                 placeholder="Enter text prompt here (optional)",
                 lines=2
             )
             
             with gr.Group():
-                with gr.Row(equal_height=True):
-                    with gr.Column(scale=1):
-                        guidance = gr.Slider(
-                            minimum=1,
-                            maximum=10,
-                            value=3.5,
-                            step=0.5,
-                            label="Guidance Scale",
-                            info="Higher values follow prompt more closely"
-                        )
-                    with gr.Column(scale=1):
-                        steps = gr.Slider(
-                            minimum=1,
-                            maximum=50,
-                            value=28,
-                            step=1,
-                            label="Steps",
-                            info="More steps = better quality but slower"
-                        )
-                
-                with gr.Row(equal_height=True):
-                    with gr.Column(scale=1):
-                        num_images = gr.Slider(
-                            minimum=1,
-                            maximum=4,
-                            value=2,
-                            step=1,
-                            label="Number of Images",
-                            info="Generate multiple variations"
-                        )
-                    with gr.Column(scale=1):
-                        seed = gr.Number(
-                            label="Random Seed",
-                            value=None,
-                            precision=0,
-                            info="Optional, for reproducibility"
-                        )
+                guidance = gr.Slider(
+                    minimum=1,
+                    maximum=10,
+                    value=3.5,
+                    step=0.5,
+                    label="Guidance Scale"
+                )
+                steps = gr.Slider(
+                    minimum=1,
+                    maximum=50,
+                    value=28,
+                    step=1,
+                    label="Number of Steps"
+                )
+                num_images = gr.Slider(
+                    minimum=1,
+                    maximum=4,
+                    value=2,
+                    step=1,
+                    label="Number of Images"
+                )
+                seed = gr.Number(
+                    label="Random Seed (optional)",
+                    precision=0
+                )
             
-            submit_btn = gr.Button(
-                "Generate Variations",
-                variant="primary",
-                scale=1
-            )
+            submit_btn = gr.Button("Generate Variations", variant="primary")
             
         with gr.Column(scale=1):
             output_gallery = gr.Gallery(
                 label="Generated Variations",
                 columns=2,
-                rows=2,
-                height=768,
-                object_fit="contain",
                 show_label=True
             )
     
-    gr.Markdown("""
-    ### Tips:
-    - Upload any image to get started
-    - Add a text prompt to guide the generation in a specific direction
-    - Adjust guidance scale to control how closely the output follows the prompt
-    - Increase steps for higher quality (but slower) generation
-    - Use the same seed to reproduce results
-    """)
-
     # Set up the generation function
     submit_btn.click(
         fn=interface.generate,
@@ -296,6 +232,5 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         outputs=output_gallery
     )
 
-# Launch the app
 if __name__ == "__main__":
     demo.launch()