Update app.py

app.py

@@ -49,7 +49,7 @@ def get_transform(resolution):
 
 # Image-to-Text Function
 @spaces.GPU
-def image_to_text(image, prompt, resolution=1024, steps=64, cfg=9.0):
+def image_to_text(image, prompt, seed=42, steps=64, cfg=9.0):
     try:
         transform = get_transform(resolution)
         
@@ -69,7 +69,7 @@ def image_to_text(image, prompt, resolution=1024, steps=64, cfg=9.0):
             guidance_scale=cfg,
             num_inference_steps=steps,
             mask_token_embedding="./mask_token_embedding.pth",
-            generator=torch.manual_seed(42),
+            generator=torch.manual_seed(seed),
         )
 
         return output.prompts[0]
@@ -79,7 +79,7 @@ def image_to_text(image, prompt, resolution=1024, steps=64, cfg=9.0):
 
 # Text-to-Image Function
 @spaces.GPU
-def text_to_image(prompt, negative_prompt, num_images=1, resolution=1024, steps=64, cfg=9.0):
+def text_to_image(prompt, negative_prompt, num_images=1, seed=42, steps=64, cfg=9.0):
     try:
         negative_prompt = negative_prompt or "worst quality, low quality, low res, blurry, distortion, watermark, logo, signature, text, jpeg artifacts, signature, sketch, duplicate, ugly, identifying mark"
         
@@ -91,7 +91,7 @@ def text_to_image(prompt, negative_prompt, num_images=1, resolution=1024, steps=
             guidance_scale=cfg,
             num_inference_steps=steps,
             mask_token_embedding="./mask_token_embedding.pth",
-            generator=torch.manual_seed(42),
+            generator=torch.manual_seed(seed),
         )
         
         return output.images
@@ -112,7 +112,7 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Muddit Unifined Model") as demo:
                 i2t_prompt_input = gr.Textbox(label="Prompt", value="Please describe this image.", placeholder="Enter your prompt here...")
                 
                 with gr.Accordion("Advanced Settings", open=False):
-                    i2t_resolution = gr.Slider(label="Resolution", minimum=256, maximum=1024, value=1024, step=64)
+                    seed = gr.Slider(label="Seed", minimum=0, maximum=2**32 - 1, step=1, value=42)
                     i2t_steps = gr.Slider(label="Inference Steps", minimum=10, maximum=100, value=64, step=1)
                     i2t_cfg = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=20.0, value=9.0, step=0.5)
                 
@@ -141,7 +141,7 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Muddit Unifined Model") as demo:
                 vqa_prompt_input = gr.Textbox(label="Prompt", placeholder="Enter your question here...")
                 
                 with gr.Accordion("Advanced Settings", open=False):
-                    vqa_resolution = gr.Slider(label="Resolution", minimum=256, maximum=1024, value=1024, step=64)
+                    seed = gr.Slider(label="Seed", minimum=0, maximum=2**32 - 1, step=1, value=42)
                     vqa_steps = gr.Slider(label="Inference Steps", minimum=10, maximum=100, value=64, step=1)
                     vqa_cfg = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=20.0, value=9.0, step=0.5)
                 
@@ -174,7 +174,7 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Muddit Unifined Model") as demo:
                 t2i_num_images = gr.Slider(label="Number of Images", minimum=1, maximum=4, value=1, step=1)
                 
                 with gr.Accordion("Advanced Settings", open=False):
-                    t2i_resolution = gr.Slider(label="Resolution", minimum=256, maximum=1024, value=1024, step=64)
+                    seed = gr.Slider(label="Seed", minimum=0, maximum=2**32 - 1, step=1, value=42)
                     t2i_steps = gr.Slider(label="Inference Steps", minimum=10, maximum=100, value=64, step=1)
                     t2i_cfg = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=20.0, value=9.0, step=0.5)
                 
@@ -199,19 +199,19 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Muddit Unifined Model") as demo:
     # Event handlers
     i2t_submit_btn.click(
         fn=image_to_text,
-        inputs=[i2t_image_input, i2t_prompt_input, i2t_resolution, i2t_steps, i2t_cfg],
+        inputs=[i2t_image_input, i2t_prompt_input, seed, i2t_steps, i2t_cfg],
         outputs=i2t_output_text
     )
     
     vqa_submit_btn.click(
         fn=image_to_text,
-        inputs=[vqa_image_input, vqa_prompt_input, vqa_resolution, vqa_steps, vqa_cfg],
+        inputs=[vqa_image_input, vqa_prompt_input, seed, vqa_steps, vqa_cfg],
         outputs=vqa_output_text
     )
     
     t2i_submit_btn.click(
         fn=text_to_image,
-        inputs=[t2i_prompt_input, t2i_negative_prompt, t2i_num_images, t2i_resolution, t2i_steps, t2i_cfg],
+        inputs=[t2i_prompt_input, t2i_negative_prompt, t2i_num_images, seed, t2i_steps, t2i_cfg],
         outputs=t2i_gallery
     )