ss

app.py

@@ -20,6 +20,15 @@ import time
 import faiss
 import re
 
+# Fix for older PyTorch versions that don't have get_default_device
+if not hasattr(torch, 'get_default_device'):
+    def get_default_device():
+        if torch.cuda.is_available():
+            return torch.device('cuda')
+        else:
+            return torch.device('cpu')
+    torch.get_default_device = get_default_device
+
 # Download NLTK resources
 try:
     nltk.data.find('tokenizers/punkt')
@@ -85,35 +94,57 @@ if 'explanations_generated' not in st.session_state:
     st.session_state.explanations_generated = False
 if 'current_job_description' not in st.session_state:
     st.session_state.current_job_description = ""
-if 'qwen3_tokenizer' not in st.session_state:
-    st.session_state.qwen3_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B")
-if 'qwen3_model' not in st.session_state:
-    st.session_state.qwen3_model = AutoModelForCausalLM.from_pretrained(
-        "Qwen/Qwen3-14B",
-        torch_dtype="auto",
-        device_map="auto"
-    )
+# Load Qwen models with error handling
+try:
+    if 'qwen3_tokenizer' not in st.session_state:
+        st.session_state.qwen3_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B")
+    if 'qwen3_model' not in st.session_state:
+        st.session_state.qwen3_model = AutoModelForCausalLM.from_pretrained(
+            "Qwen/Qwen3-14B",
+            torch_dtype="auto",
+            device_map="auto"
+        )
+except Exception as e:
+    st.warning(f"⚠️ Could not load Qwen3-14B: {str(e)}")
+    st.session_state.qwen3_tokenizer = None
+    st.session_state.qwen3_model = None
+
 # Separate smaller model for intent analysis
-if 'qwen3_intent_tokenizer' not in st.session_state:
-    st.session_state.qwen3_intent_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-1.7B")
-if 'qwen3_intent_model' not in st.session_state:
-    st.session_state.qwen3_intent_model = AutoModelForCausalLM.from_pretrained(
-        "Qwen/Qwen3-1.7B",
-        torch_dtype="auto",
-        device_map="auto"
-    )
+try:
+    if 'qwen3_intent_tokenizer' not in st.session_state:
+        st.session_state.qwen3_intent_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-1.7B")
+    if 'qwen3_intent_model' not in st.session_state:
+        st.session_state.qwen3_intent_model = AutoModelForCausalLM.from_pretrained(
+            "Qwen/Qwen3-1.7B",
+            torch_dtype="auto",
+            device_map="auto"
+        )
+except Exception as e:
+    st.warning(f"⚠️ Could not load Qwen3-1.7B: {str(e)}")
+    st.session_state.qwen3_intent_tokenizer = None
+    st.session_state.qwen3_intent_model = None
 
 @st.cache_resource
 def load_embedding_model():
     """Load and cache the BGE embedding model"""
     try:
         with st.spinner("🔄 Loading BAAI/bge-large-en-v1.5 model..."):
-            model = SentenceTransformer('BAAI/bge-large-en-v1.5')
+            # Try with explicit device specification
+            device = 'cuda' if torch.cuda.is_available() else 'cpu'
+            model = SentenceTransformer('BAAI/bge-large-en-v1.5', device=device)
             st.success("✅ Embedding model loaded successfully!")
             return model
     except Exception as e:
         st.error(f"❌ Error loading embedding model: {str(e)}")
-        return None
+        try:
+            # Fallback: try with a smaller model
+            st.warning("🔄 Trying fallback model: all-MiniLM-L6-v2...")
+            model = SentenceTransformer('all-MiniLM-L6-v2')
+            st.success("✅ Fallback embedding model loaded!")
+            return model
+        except Exception as e2:
+            st.error(f"❌ Fallback also failed: {str(e2)}")
+            return None
 
 @st.cache_resource
 def load_cross_encoder():
@@ -121,12 +152,22 @@ def load_cross_encoder():
     try:
         with st.spinner("🔄 Loading Cross-Encoder ms-marco-MiniLM-L6-v2..."):
             from sentence_transformers import CrossEncoder
-            model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L6-v2')
+            # Try with explicit device specification
+            device = 'cuda' if torch.cuda.is_available() else 'cpu'
+            model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L6-v2', device=device)
             st.success("✅ Cross-Encoder model loaded successfully!")
             return model
     except Exception as e:
         st.error(f"❌ Error loading Cross-Encoder model: {str(e)}")
-        return None
+        try:
+            # Fallback: try without device specification
+            st.warning("🔄 Trying Cross-Encoder without device specification...")
+            model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L6-v2')
+            st.success("✅ Cross-Encoder model loaded (fallback)!")
+            return model
+        except Exception as e2:
+            st.error(f"❌ Cross-Encoder fallback also failed: {str(e2)}")
+            return None
 
 def generate_qwen3_response(prompt, tokenizer, model, max_new_tokens=200):
     messages = [{"role": "user", "content": prompt}]

requirements.txt

@@ -1,6 +1,6 @@
 streamlit==1.31.0
 transformers>=4.51.0
-torch==2.1.2
+torch>=2.1.0
 pdfplumber==0.10.1
 PyPDF2==3.0.1
 python-docx==1.0.1