Update chatbot_rag.py

chatbot_rag.py

@@ -39,17 +39,14 @@ def build_qa():
     print("📂 Docs in DB:", vectorstore._collection.count())
 
     # 3. Load LLM (Phi-3 mini)
-    print("🔹 Loading LLM...")
+    model_id = "microsoft/VibeVoice-1.5B"
 
-    model_id = "microsoft/VibeVoice-1.5B" 
-    
     tokenizer = AutoTokenizer.from_pretrained(model_id)
-    
     model = AutoModelForSeq2SeqLM.from_pretrained(
         model_id,
-        device_map="auto",       # use available CPU/GPU
-        torch_dtype="auto",      # lets HF decide best precision
-        trust_remote_code=True   # Gemma needs this
+        device_map="auto",
+        torch_dtype="auto",
+        trust_remote_code=True
     )
     
     pipe = pipeline(
@@ -57,16 +54,17 @@ def build_qa():
         model=model,
         tokenizer=tokenizer,
         max_new_tokens=256,
-        temperature=0.2,         # mostly deterministic
-        do_sample=True,          # allow some controlled randomness
-        top_p=0.9,               # nucleus sampling
-        repetition_penalty=1.2,  # 🚀 avoid loops/repeats
+        temperature=0.2,
+        do_sample=True,
+        top_p=0.9,
+        repetition_penalty=1.2,
         eos_token_id=tokenizer.eos_token_id,
         return_full_text=False
     )
     
     llm = HuggingFacePipeline(pipeline=pipe)
 
+
     # 4. Retriever
     retriever = vectorstore.as_retriever(search_kwargs={"k": 3})