Fix: remove gr.Progress (breaks API schema in Gradio 5.20, kills all inference)

app.py

@@ -1,6 +1,5 @@
 """
 Virtual Try-On — Paint-by-Example + Hugging Face ZeroGPU
-Uses an exemplar garment image to guide inpainting on a person photo.
 No local GPU or model storage needed.
 """
 
@@ -13,7 +12,7 @@ import torch
 from PIL import Image, ImageDraw
 
 # ---------------------------------------------------------------------------
-# Persistent storage (/data on ZeroGPU Spaces, /tmp fallback)
+# Persistent storage
 # ---------------------------------------------------------------------------
 DATA_DIR   = "/data" if os.path.exists("/data") else "/tmp"
 OUTPUT_DIR = os.path.join(DATA_DIR, "outputs")
@@ -46,13 +45,22 @@ def _make_mask(size: int, cloth_type: str) -> Image.Image:
     return mask
 
 # ---------------------------------------------------------------------------
-# GPU-decorated inference
+# GPU inference
 # ---------------------------------------------------------------------------
 _pipe = None
 
 @spaces.GPU(duration=120)
-def _run_inference(person: Image.Image, garment: Image.Image, mask: Image.Image,
-                   num_steps: int, guidance_scale: float, seed: int):
+def run_tryon(
+    person_image: Image.Image,
+    garment_image: Image.Image,
+    cloth_type: str,
+    num_steps: int,
+    guidance_scale: float,
+    seed: int,
+):
+    if person_image is None or garment_image is None:
+        raise gr.Error("Please upload both a person photo and a garment image.")
+
     global _pipe
     if _pipe is None:
         from diffusers import PaintByExamplePipeline
@@ -64,6 +72,10 @@ def _run_inference(person: Image.Image, garment: Image.Image, mask: Image.Image,
         _pipe.set_progress_bar_config(disable=True)
         print("Pipeline ready.")
 
+    person  = _fit_to_square(person_image)
+    garment = _fit_to_square(garment_image)
+    mask    = _make_mask(TARGET_SIZE, cloth_type)
+
     rng = torch.Generator(device="cuda")
     rng.manual_seed(int(seed) if seed != -1 else torch.randint(0, 2**32, (1,)).item())
 
@@ -75,32 +87,7 @@ def _run_inference(person: Image.Image, garment: Image.Image, mask: Image.Image,
         guidance_scale=guidance_scale,
         generator=rng,
     )
-    return result.images
-
-# ---------------------------------------------------------------------------
-# Gradio inference — uses gr.Progress for live status updates
-# ---------------------------------------------------------------------------
-def run_tryon(
-    person_image: Image.Image,
-    garment_image: Image.Image,
-    cloth_type: str,
-    num_steps: int,
-    guidance_scale: float,
-    seed: int,
-    progress=gr.Progress(track_tqdm=True),
-):
-    if person_image is None or garment_image is None:
-        raise gr.Error("Please upload both a person photo and a garment image.")
-
-    progress(0, desc="⏳ Requesting GPU + loading model (first run ~3 min)…")
-
-    person  = _fit_to_square(person_image)
-    garment = _fit_to_square(garment_image)
-    mask    = _make_mask(TARGET_SIZE, cloth_type)
-
-    output_images = _run_inference(person, garment, mask, num_steps, guidance_scale, seed)
-
-    progress(0.9, desc="💾 Saving result…")
+    output_images = result.images
 
     timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
     saved_paths = []
@@ -109,8 +96,7 @@ def run_tryon(
         img.save(path, format="PNG")
         saved_paths.append(path)
 
-    progress(1.0, desc="✅ Done!")
-    return output_images, saved_paths, "✅ Done! Download your result below."
+    return output_images, saved_paths
 
 # ---------------------------------------------------------------------------
 # Gradio UI
@@ -119,8 +105,9 @@ with gr.Blocks(title="Virtual Try-On", theme=gr.themes.Soft()) as demo:
     gr.Markdown(
         "# 👗 Virtual Try-On\n"
         "Upload a **person photo** and a **garment image**, select the type, then click **Try On**.\n\n"
-        "> Runs on **Hugging Face ZeroGPU** (free A10G) — no local GPU needed.  \n"
-        "> **First run:** ~2-3 min (model download ~5 GB). **Subsequent runs:** ~15-30s."
+        "> Runs on **Hugging Face ZeroGPU** (free A10G) — no local GPU needed.\n\n"
+        "> ⏳ After clicking, the button shows a **loading spinner** — that means it is working.  \n"
+        "> **First run:** ~2-3 min (model download). **Subsequent runs:** ~15-30s."
     )
 
     with gr.Row():
@@ -140,7 +127,6 @@ with gr.Blocks(title="Virtual Try-On", theme=gr.themes.Soft()) as demo:
             try_btn = gr.Button("👗 Try On", variant="primary", size="lg")
 
         with gr.Column():
-            status_text    = gr.Textbox(label="Status", value="Ready", interactive=False, max_lines=1)
             output_gallery = gr.Gallery(label="Result", columns=1, height=380)
             output_files   = gr.File(
                 label="⬇ Download to your device",
@@ -151,7 +137,7 @@ with gr.Blocks(title="Virtual Try-On", theme=gr.themes.Soft()) as demo:
     try_btn.click(
         fn=run_tryon,
         inputs=[person_input, garment_input, cloth_type, num_steps, guidance, seed_input],
-        outputs=[output_gallery, output_files, status_text],
+        outputs=[output_gallery, output_files],
     )
 
     gr.Markdown(