Update app.py

app.py

@@ -2,22 +2,27 @@ import io
 import base64
 import torch
 from flask import Flask, request, jsonify
-from diffusers import StableDiffusionPipeline  # Placeholder; SF3D uses a custom setup
+from diffusers import StableDiffusionPipeline  # Placeholder; adjust based on SF3D docs
 from PIL import Image
 import logging
+import os
+from huggingface_hub import HfFolder
 
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 app = Flask(__name__)
 
-# Load the model once  startup (on CPU)
+# Load the model once at startup (on CPU) with authentication
 try:
     logger.info("Loading Stable Fast 3D pipeline...")
+    # Use your Hugging Face token (set as environment variable or hardcoded for testing)
+    token = os.getenv("HF_TOKEN") or "your_hf_token_here"  # Replace with your token
     pipe = StableDiffusionPipeline.from_pretrained(
         "stabilityai/stable-fast-3d",
         torch_dtype=torch.float32,
         cache_dir="/tmp/hf_home",
+        token=token,  # Pass token for authentication
     )
     pipe.to("cpu")
     logger.info("=== Application Startup at CPU mode =====")
@@ -52,11 +57,9 @@ def generate():
         image = Image.open(io.BytesIO(base64.b64decode(image_data))).convert("RGB")
 
         logger.info("Processing image with pipeline...")
-        # SF3D-specific generation; adjust based on documentation
-        result = pipe(image)  # Placeholder; check SF3D repo for exact method
-        output_mesh = result.mesh  # Hypothetical output; may be .glb or .obj
+        result = pipe(image)  # Adjust based on SF3D documentation
+        output_mesh = result.mesh  # Hypothetical; check SF3D output format
 
-        # Save mesh to temporary file and encode
         output_path = "/tmp/output.glb"
         output_mesh.save(output_path)
         with open(output_path, "rb") as f: