Enhance model loading function with connectivity checks and detailed error reporting

utils/prediction.py

@@ -3,13 +3,24 @@ from transformers import AutoTokenizer
 from sklearn.preprocessing import LabelEncoder
 from utils.BiLSTM import BiLSTMAttentionBERT
 import numpy as np
+import streamlit as st
+import requests
 
 
 
 def load_model_for_prediction():
-    
     try:
-        # Load model from Hugging Face Hub
+        st.write("Starting model loading...")
+        
+        # Test Hugging Face connectivity
+        st.write("Testing connection to Hugging Face...")
+        response = requests.get("https://huggingface.co/joko333/BiLSTM_v01")
+        if response.status_code != 200:
+            st.error(f"Cannot connect to Hugging Face. Status code: {response.status_code}")
+            return None, None, None
+            
+        # Load model with logging
+        st.write("Loading BiLSTM model...")
         model = BiLSTMAttentionBERT.from_pretrained(
             "joko333/BiLSTM_v01",
             hidden_dim=128,
@@ -17,10 +28,10 @@ def load_model_for_prediction():
             num_layers=2,
             dropout=0.5
         )
+        st.write("Model loaded successfully")
         
-        model.eval()
-        
-        # Initialize label encoder with predefined classes
+        # Initialize label encoder
+        st.write("Initializing label encoder...")
         label_encoder = LabelEncoder()
         label_encoder.classes_ = np.array(['Addition', 'Causal', 'Cause and Effect', 
                                          'Clarification', 'Comparison', 'Concession', 
@@ -30,16 +41,20 @@ def load_model_for_prediction():
                                          'Illustration', 'Inference', 'Problem Solution', 
                                          'Purpose', 'Sequential', 'Summary', 
                                          'Temporal Sequence'])
+        st.write("Label encoder initialized")
         
-        # Initialize tokenizer
-        tokenizer = AutoTokenizer.from_pretrained(
-            'dmis-lab/biobert-base-cased-v1.2'
-        )
+        # Load tokenizer
+        st.write("Loading tokenizer...")
+        tokenizer = AutoTokenizer.from_pretrained('dmis-lab/biobert-base-cased-v1.2')
+        st.write("Tokenizer loaded successfully")
         
         return model, label_encoder, tokenizer
         
     except Exception as e:
-        print(f"Error loading model components: {str(e)}")
+        st.error(f"Detailed error: {str(e)}")
+        st.error(f"Error type: {type(e).__name__}")
+        import traceback
+        st.error(f"Traceback: {traceback.format_exc()}")
         return None, None, None
 
 def predict_sentence(model, sentence, tokenizer, label_encoder):