Update app.py

app.py

@@ -9,6 +9,21 @@ from sentence_transformers import SentenceTransformer, util
 import numpy as np
 from transformers import pipeline, AutoModelForQuestionAnswering, AutoTokenizer
 import torch
+from huggingface_hub import hf_hub_download, HfFolder
+from huggingface_hub.utils import configure_http_backend
+import requests as hf_requests
+
+# Configure Hugging Face Hub to use a custom session with increased timeout and retries
+def create_custom_session():
+    session = hf_requests.Session()
+    # Increase timeout to 30 seconds (default is 10 seconds)
+    adapter = hf_requests.adapters.HTTPAdapter(max_retries=3)  # Retry 3 times on failure
+    session.mount("https://", adapter)
+    session.timeout = 30  # Set timeout to 30 seconds
+    return session
+
+# Set the custom session for Hugging Face Hub
+configure_http_backend(backend_factory=create_custom_session)
 
 # Global variables for in-memory storage (reset on app restart)
 corpus = []  # List of paragraphs from URLs
@@ -21,8 +36,14 @@ retriever = SentenceTransformer('all-MiniLM-L6-v2')
 
 # Load PyTorch model for QA
 # Model: distilbert-base-uncased-distilled-squad (~260MB)
-model = AutoModelForQuestionAnswering.from_pretrained("distilbert-base-uncased-distilled-squad")
-tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased-distilled-squad")
+try:
+    model = AutoModelForQuestionAnswering.from_pretrained("distilbert-base-uncased-distilled-squad")
+    tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased-distilled-squad")
+except Exception as e:
+    print(f"Error loading model: {str(e)}. Retrying with force_download=True...")
+    # Force re-download in case of corrupted cache
+    model = AutoModelForQuestionAnswering.from_pretrained("distilbert-base-uncased-distilled-squad", force_download=True)
+    tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased-distilled-squad", force_download=True)
 
 # Set model to evaluation mode
 model.eval()