Update app.py

app.py

@@ -160,58 +160,50 @@ def generate_image_to_image(prompt_mash, image_input_path, image_strength, steps
     return final_image 
 
 @spaces.GPU(duration=70)
-def run_lora(prompt, image_input, image_strength, cfg_scale, steps, selected_indices_json, randomize_seed, seed, width, height, lora_scale, progress=gr.Progress(track_tqdm=True)):
+def run_lora(prompt, image_input, image_strength, cfg_scale, steps, selected_indices_json, selected_weights_json, randomize_seed, seed, width, height, global_lora_scale, progress=gr.Progress(track_tqdm=True)):
     import json
-    # Parse the JSON string into a list of indices
+    # Parse the JSON strings
     selected_indices = json.loads(selected_indices_json)
+    selected_weights = json.loads(selected_weights_json) if selected_weights_json else {}
     
-    # Ensure at least one LoRA is selected
-    if not selected_indices or len(selected_indices) == 0:
+    if not selected_indices:
         raise gr.Error("You must select at least one LoRA before proceeding.")
     
-    # Combine trigger words from all selected LoRAs into the prompt
+    # Combine trigger words from all selected LoRAs
     prompt_mash = prompt
     for idx in selected_indices:
         selected_lora = loras[idx]
         if "trigger_word" in selected_lora and selected_lora["trigger_word"]:
-            # Prepend each trigger word to the prompt; you can adjust the order or separator as needed
             prompt_mash = f"{selected_lora['trigger_word']} {prompt_mash}"
     
-    # Unload any previously loaded LoRA weights
     with calculateDuration("Unloading LoRA"):
         pipe.unload_lora_weights()
         pipe_i2i.unload_lora_weights()
     
-    # Load each selected LoRA weight sequentially
     with calculateDuration("Loading LoRA weights"):
-        # Use the image-to-image pipeline if an input image is provided, else the text-to-image pipeline
         pipe_to_use = pipe_i2i if image_input is not None else pipe
         for idx in selected_indices:
             selected_lora = loras[idx]
             weight_name = selected_lora.get("weights", None)
+            # Get the individual weight for this LoRA from the selected_weights mapping.
+            # If not found, default to 0.95.
+            lora_weight = selected_weights.get(str(idx), 0.95)
             pipe_to_use.load_lora_weights(
-                selected_lora["repo"], 
-                weight_name=weight_name, 
-                low_cpu_mem_usage=True
+                selected_lora["repo"],
+                weight_name=weight_name,
+                low_cpu_mem_usage=True,
+                lora_weight=lora_weight  # This parameter should be supported by your load function.
             )
     
-    # Optionally randomize the seed if requested
     with calculateDuration("Randomizing seed"):
         if randomize_seed:
-            seed = random.randint(0, MAX_SEED)
+            seed = random.randint(0, 2**32-1)
     
-    # Generate image(s)
     if image_input is not None:
-        # Image-to-image generation
-        final_image = generate_image_to_image(
-            prompt_mash, image_input, image_strength, steps, cfg_scale, width, height, lora_scale, seed
-        )
+        final_image = generate_image_to_image(prompt_mash, image_input, image_strength, steps, cfg_scale, width, height, global_lora_scale, seed)
         yield final_image, seed, gr.update(visible=False)
     else:
-        # Text-to-image generation
-        image_generator = generate_image(
-            prompt_mash, steps, seed, cfg_scale, width, height, lora_scale, progress
-        )
+        image_generator = generate_image(prompt_mash, steps, seed, cfg_scale, width, height, global_lora_scale, progress)
         final_image = None
         step_counter = 0
         for image in image_generator:
@@ -222,6 +214,7 @@ def run_lora(prompt, image_input, image_strength, cfg_scale, steps, selected_ind
         yield final_image, seed, gr.update(value=progress_bar, visible=False)
 
 
+
         
 def get_huggingface_safetensors(link):
   split_link = link.split("/")
@@ -301,18 +294,18 @@ def remove_custom_lora():
 run_lora.zerogpu = True
 
 css = '''
-#gen_btn{height: 100%}
-#gen_column{align-self: stretch}
-#title{text-align: center}
-#title h1{font-size: 3em; display:inline-flex; align-items:center}
-#title img{width: 100px; margin-right: 0.5em}
-#lora_list{background: var(--block-background-fill);padding: 0 1em .3em; font-size: 90%}
-.card_internal{display: flex;height: 100px;margin-top: .5em}
-.card_internal img{margin-right: 1em}
-.styler{--form-gap-width: 0px !important}
-#progress{height:30px}
-.progress-container {width: 100%;height: 30px;background-color: #f0f0f0;border-radius: 15px;overflow: hidden;margin-bottom: 20px}
-.progress-bar {height: 100%;background-color: #4f46e5;width: calc(var(--current) / var(--total) * 100%);transition: width 0.5s ease-in-out}
+#gen_btn { height: 100%; }
+#gen_column { align-self: stretch; }
+#title { text-align: center; }
+#title h1 { font-size: 3em; display: inline-flex; align-items: center; }
+#title img { width: 100px; margin-right: 0.5em; }
+#lora_list { background: var(--block-background-fill); padding: 0 1em .3em; font-size: 90%; }
+.card_internal { display: flex; height: 100px; margin-top: .5em; }
+.card_internal img { margin-right: 1em; }
+.styler { --form-gap-width: 0px !important; }
+#progress { height: 30px; }
+.progress-container { width: 100%; height: 30px; background-color: #f0f0f0; border-radius: 15px; overflow: hidden; margin-bottom: 20px; }
+.progress-bar { height: 100%; background-color: #4f46e5; width: calc(var(--current) / var(--total) * 100%); transition: width 0.5s ease-in-out; }
 '''
 font = [gr.themes.GoogleFont("Source Sans Pro"), "Arial", "sans-serif"]
 
@@ -321,8 +314,10 @@ with gr.Blocks(theme=gr.themes.Soft(font=font), css=css, delete_cache=(60, 60))
         """<h1><img src="https://huggingface.co/spaces/kayte0342/test/resolve/main/DA4BE61E-A0BD-4254-A1B6-AD3C05D18A9C%20(1).png?download=true" alt="LoRA"> FLUX LoRA Kayte's Space</h1>""",
         elem_id="title",
     )
-    # Hidden textbox to store the JSON string of selected indices
+    
+    # Hidden textboxes to store the JSON outputs:
     selected_indices_hidden = gr.Textbox(value="[]", visible=False)
+    selected_weights_hidden = gr.Textbox(value="{}", visible=False)
     
     with gr.Row():
         with gr.Column(scale=3):
@@ -333,15 +328,19 @@ with gr.Blocks(theme=gr.themes.Soft(font=font), css=css, delete_cache=(60, 60))
     with gr.Row():
         with gr.Column():
             selected_info = gr.Markdown("")
-            # Build a custom layout for LoRA selection with checkboxes.
+            # Build a custom layout for LoRA selection.
             lora_selection_container = gr.Column()
+            # We'll collect checkbox and slider components in lists.
             lora_checkbox_list = []
+            lora_slider_list = []
             for idx, lora in enumerate(loras):
                 with gr.Row():
-                    gr.Image(value=lora["image"], label=lora["title"], height=100)
+                    gr.Image(label=lora["title"], height=100)
                     checkbox = gr.Checkbox(label="Select", value=False, elem_id=f"lora_checkbox_{idx}")
+                    slider = gr.Slider(label="Weight", minimum=0, maximum=3, step=0.01, value=0.95, elem_id=f"lora_weight_{idx}")
                     lora_checkbox_list.append(checkbox)
-            gr.Markdown("[Check the list of FLUX LoRas](https://huggingface.co/models?other=base_model:adapter:black-forest-labs/FLUX.1-dev)", elem_id="lora_list")
+                    lora_slider_list.append(slider)
+            gr.Markdown("[Check the list of FLUX LoRAs](https://huggingface.co/models?other=base_model:adapter:black-forest-labs/FLUX.1-dev)", elem_id="lora_list")
         with gr.Column():
             progress_bar = gr.Markdown(elem_id="progress", visible=False)
             result = gr.Image(label="Generated Image")
@@ -361,27 +360,35 @@ with gr.Blocks(theme=gr.themes.Soft(font=font), css=css, delete_cache=(60, 60))
                 with gr.Row():
                     randomize_seed = gr.Checkbox(True, label="Randomize seed")
                     seed = gr.Slider(label="Seed", minimum=0, maximum=2**32-1, step=1, value=0, randomize=True)
-                    lora_scale = gr.Slider(label="LoRA Scale", minimum=0, maximum=3, step=0.01, value=0.95)
+                    lora_scale = gr.Slider(label="Global LoRA Scale", minimum=0, maximum=3, step=0.01, value=0.95)
     
-    # Function to combine the checkbox values into a JSON string.
+    # Function to combine checkbox selections into a JSON list of indices.
     def combine_selections(*checkbox_values):
         selected_indices = [i for i, v in enumerate(checkbox_values) if v]
         return json.dumps(selected_indices)
     
-    # Chain the update: When the Generate button is clicked,
-    # first update the hidden state with combine_selections,
-    # then run run_lora using the updated hidden state.
+    # Function to combine all slider values into a JSON dictionary mapping index to weight.
+    def combine_weights(*slider_values):
+        weights = {str(i): v for i, v in enumerate(slider_values)}
+        return json.dumps(weights)
+    
+    # Chain the updates when the Generate button is clicked:
+    # First, update the checkbox hidden state, then update the slider hidden state, then call run_lora.
     generate_button.click(
         combine_selections,
         inputs=lora_checkbox_list,
         outputs=selected_indices_hidden
+    ).then(
+        combine_weights,
+        inputs=lora_slider_list,
+        outputs=selected_weights_hidden
     ).then(
         run_lora,
-        inputs=[prompt, input_image, image_strength, cfg_scale, steps, selected_indices_hidden, randomize_seed, seed, width, height, lora_scale],
+        inputs=[prompt, input_image, image_strength, cfg_scale, steps, selected_indices_hidden, selected_weights_hidden, randomize_seed, seed, width, height, lora_scale],
         outputs=[result, seed, progress_bar]
     )
     
-    # Optionally, update the selected_info markdown when the hidden state changes.
+    # Update the selected_info display when the selected_indices_hidden changes.
     def update_info(selected_json):
         selected_indices = json.loads(selected_json)
         if selected_indices: