Update chatbot_rag.py

chatbot_rag.py

@@ -39,12 +39,16 @@ def build_qa():
 
     # 3. Load LLM (Phi-3 mini)
     print("🔹 Loading LLM...")
-    model_id = "openai-community/gpt2-xl"   # ~1.5B GPT-2
+
+    model_id = "google/gemma-2-2b-it" 
     
     tokenizer = AutoTokenizer.from_pretrained(model_id)
+    
     model = AutoModelForCausalLM.from_pretrained(
         model_id,
-        device_map="auto"
+        device_map="auto",       # use available CPU/GPU
+        torch_dtype="auto",      # lets HF decide best precision
+        trust_remote_code=True   # Gemma needs this
     )
     
     pipe = pipeline(
@@ -52,9 +56,11 @@ def build_qa():
         model=model,
         tokenizer=tokenizer,
         max_new_tokens=256,
-        temperature=0.2,
-        do_sample=False,
-        repetition_penalty=1.2,
+        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
+        eos_token_id=tokenizer.eos_token_id,
         return_full_text=False
     )