Update app.py

app.py

@@ -1,6 +1,9 @@
 # -*- coding: utf-8 -*-
 """
-Author: @Raxephion 2025
+Hugging Face Spaces Script for Basic Stable Diffusion 1.5 Gradio App
+Adapted from user's local script and HF Spaces template.
+Supports Hub models and CPU/GPU selection based on available hardware.
+Includes Attention Slicing optimization toggle.
 """
 
 import gradio as gr
@@ -102,10 +105,15 @@ if INITIAL_MODEL_ID:
         )
 
         # --- Apply Optimizations during initial load ---
-        # Apply attention slicing by default for memory efficiency on Spaces
-        # Can be turned off via UI toggle later, but good default for VRAM
-        # We'll add the UI toggle later, for now, just enable it here
-        # pipeline.enable_attention_slicing() # Enable by default on initial load
+        # Apply attention slicing by default if GPU is available for memory efficiency on Spaces
+        if initial_device_to_use == "cuda":
+            try:
+                pipeline.enable_attention_slicing()
+                print("Attention Slicing enabled during initial load.")
+            except Exception as e:
+                print(f"Warning: Failed to enable Attention Slicing during initial load: {e}")
+                # Don't raise Gradio error here, just print warning
+
 
         pipeline = pipeline.to(initial_device_to_use) # Move to the initial device
 
@@ -155,8 +163,9 @@ def infer(
     size,                     # From size_dropdown
     seed,                     # From seed_input (now a Slider)
     randomize_seed,           # From randomize_seed_checkbox
-    enable_attention_slicing, # <-- New input for the optimization toggle
-    progress=gr.Progress(track_tqdm=True), # Added progress argument from template
+    enable_attention_slicing, # New input for the optimization toggle
+    # Removed track_tqdm=True which can conflict with callbacks
+    progress=gr.Progress(),   # <-- Corrected Progress initialization
 ):
     """Generates an image using the selected model and parameters on the chosen device."""
     global current_pipeline, current_model_id, current_device_loaded, SCHEDULER_MAP, MAX_SEED
@@ -213,20 +222,20 @@ def infer(
              if enable_attention_slicing and temp_device_to_use == "cuda": # Only apply on GPU
                  try:
                      pipeline.enable_attention_slicing()
-                     print("Attention Slicing enabled.")
+                     print("Attention Slicing enabled during model load.")
                  except Exception as e:
                      print(f"Warning: Failed to enable Attention Slicing: {e}")
                      gr.Warning(f"Failed to enable Attention Slicing. Error: {e}")
              else:
                  try:
                      pipeline.disable_attention_slicing() # Ensure it's off if toggle is off or on CPU
-                     print("Attention Slicing disabled.")
+                     # print("Attention Slicing disabled during model load.") # Avoid noise
                  except Exception as e:
                       # May fail if it wasn't enabled, ignore
                       pass
 
 
-             pipeline = pipeline.to(temp_device_to_use) # Use the determined device
+             pipeline = pipeline.to(temp_device_to_use) # Move to the determined device
 
              current_pipeline = pipeline
              current_model_id = model_identifier
@@ -285,6 +294,7 @@ def infer(
 
     # --- Apply Optimizations *before* generation if model was already loaded ---
     # If the model didn't need reloading, we need to apply/remove slicing here
+    # Check the pipeline's actual device
     if str(current_pipeline.device) == "cuda": # Only attempt on GPU
         if enable_attention_slicing:
              try:
@@ -374,7 +384,7 @@ def infer(
              print(f"Warning: Invalid seed input '{seed}'. Using random seed instead.")
              gr.Warning(f"Invalid seed input '{seed}'. Using random seed instead.")
              seed_int = random.randint(0, MAX_SEED) # Fallback to random if input is not int
-             randomize_seed = True # Mark as randomized for reporting
+             # No need to set randomize_seed = True here, just use the new random seed_int
 
 
     generator = None # Initialize generator as None
@@ -536,11 +546,11 @@ with gr.Blocks(css=css, theme=gr.themes.Soft()) as demo: # Added Soft theme from
                     )
                 # Combine seed input and randomize checkbox
                 with gr.Row():
-                     # Removed precision=0 from Slider
+                     # Removed precision=0 from Slider - FIX FOR TYPEERROR
                      seed_input = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=0, interactive=True) # Use 0 as default, interactive initially
                      randomize_seed_checkbox = gr.Checkbox(label="Randomize seed", value=True) # Simplified label
 
-                # --- New: Memory Optimization Toggle ---
+                # --- Memory Optimization Toggle ---
                 with gr.Row():
                      # Default to enabled if GPU is available, otherwise off
                      default_slicing = True if "GPU" in AVAILABLE_DEVICES else False
@@ -584,7 +594,7 @@ with gr.Blocks(css=css, theme=gr.themes.Soft()) as demo: # Added Soft theme from
             size_dropdown,
             seed_input,
             randomize_seed_checkbox,
-            enable_attention_slicing_checkbox, # <-- Pass the new checkbox value
+            enable_attention_slicing_checkbox, # Pass the new checkbox value
         ],
         outputs=[output_image, actual_seed_output], # Return image and the actual seed used
         api_name="generate" # Optional: For API access