Update app.py

app.py

@@ -1,10 +1,19 @@
 # app.py
+
+import os
+import sys
+
+# --- Install Dependencies ---
+print("Installing required packages: diffusers, gradio_imageslider, huggingface-hub...")
+os.system("pip install --no-input diffusers gradio_imageslider huggingface-hub")
+
+# --- Standard Imports ---
 import logging
 import random
 import warnings
-import os
 import io
 import base64
+
 import gradio as gr
 import numpy as np
 import spaces
@@ -44,9 +53,9 @@ logging.info(f"Selected device: {device} | Data type: {torch_dtype}")
 # --- Authentication and Model Download ---
 huggingface_token = os.getenv("HUGGINGFACE_TOKEN")
 
-flux_model_id    = "black-forest-labs/FLUX.1-dev"
+flux_model_id       = "black-forest-labs/FLUX.1-dev"
 controlnet_model_id = "jasperai/Flux.1-dev-Controlnet-Upscaler"
-local_model_dir = flux_model_id.split('/')[-1]
+local_model_dir     = flux_model_id.split('/')[-1]
 pipe = None
 
 try:
@@ -78,20 +87,20 @@ except Exception as e:
     logging.error(f"Model load failed: {e}", exc_info=True)
     raise SystemExit(f"Fatal: could not load models: {e}")
 
-
-# --- Constants & Helpers ---
+# --- Constants & Helper Functions ---
 MAX_SEED = 2**32 - 1
 MAX_PIXEL_BUDGET = 1280 * 1280
 INTERNAL_PROCESSING_FACTOR = 4
 
 def process_input(input_image):
     if input_image is None:
-        raise gr.Error("No input image!")
+        raise gr.Error("No input image provided!")
     input_image = ImageOps.exif_transpose(input_image)
     if input_image.mode != 'RGB':
         input_image = input_image.convert('RGB')
     w, h = input_image.size
-    # Compute intermediate budget
+
+    # Check intermediate budget
     target_pixels = (w*INTERNAL_PROCESSING_FACTOR)*(h*INTERNAL_PROCESSING_FACTOR)
     if target_pixels > MAX_PIXEL_BUDGET:
         max_in = MAX_PIXEL_BUDGET / (INTERNAL_PROCESSING_FACTOR**2)
@@ -101,11 +110,13 @@ def process_input(input_image):
         was_resized = True
     else:
         was_resized = False
-    # round dims to multiple of 8
+
+    # Round to multiple of 8
     w2, h2 = input_image.size
-    w2 -= w2%8; h2 -= h2%8
+    w2 -= w2 % 8; h2 -= h2 % 8
     if input_image.size != (w2,h2):
         input_image = input_image.resize((w2,h2), Image.Resampling.LANCZOS)
+
     return input_image, w, h, was_resized
 
 @spaces.GPU(duration=75)
@@ -125,11 +136,13 @@ def infer(
         seed = random.randint(0, MAX_SEED)
     seed = int(seed)
     final_upscale_factor = int(final_upscale_factor)
+
     processed, w0, h0, resized_flag = process_input(input_image)
     w_proc, h_proc = processed.size
-    # control image dims
+
     cw, ch = w_proc*INTERNAL_PROCESSING_FACTOR, h_proc*INTERNAL_PROCESSING_FACTOR
     control_img = processed.resize((cw, ch), Image.Resampling.LANCZOS)
+
     generator = torch.Generator(device=device).manual_seed(seed)
     with torch.inference_mode():
         result = pipe(
@@ -141,21 +154,23 @@ def infer(
             height=ch, width=cw,
             generator=generator,
         ).images[0]
-    # final resize
+
+    # Final resize
     if resized_flag:
         final_w, final_h = w_proc*final_upscale_factor, h_proc*final_upscale_factor
     else:
         final_w, final_h = w0*final_upscale_factor, h0*final_upscale_factor
+
     if (final_w, final_h) != result.size:
         result = result.resize((final_w, final_h), Image.Resampling.LANCZOS)
-    # encode base64
+
     buf = io.BytesIO()
     result.save(buf, format="WEBP", quality=90)
     b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
-    return [[input_image, result], seed, f"data:image/webp;base64,{b64}"]
 
+    return [[input_image, result], seed, f"data:image/webp;base64,{b64}"]
 
-# --- Gradio UI ---
+# --- Gradio Interface ---
 with gr.Blocks(css=css, theme=gr.themes.Soft(), title="Flux Upscaler Demo") as demo:
     gr.Markdown(f"""
     # ⚡ Flux.1‑dev Upscaler
@@ -200,5 +215,5 @@ with gr.Blocks(css=css, theme=gr.themes.Soft(), title="Flux Upscaler Demo") as d
         api_name="upscale"
     )
 
-# expose JSON API at /run/upscale
+# Expose JSON API at /run/upscale
 demo.queue(max_size=10).launch(share=False, show_api=True)