Update src/RAGSample.py

src/RAGSample.py

@@ -18,6 +18,7 @@ import pandas as pd
 from typing import Optional, List
 import re
 import torch
+import subprocess
 
 # OPTION 1: Use Hugging Face Pipeline (Recommended for HF Spaces)
 from transformers import pipeline
@@ -387,17 +388,18 @@ Answer:
     
     # Initialize a local Hugging Face model
     hf_pipeline = pipeline(
-        "text-generation", 
-        model="Narrativaai/BioGPT-Large-finetuned-chatdoctor",
-        tokenizer="Narrativaai/BioGPT-Large-finetuned-chatdoctor",
-        max_new_tokens=100,
-        max_length=512,
-        temperature=0.3,               # Lower temp for medical accuracy
-        device="cpu",
+        "text-generation",
+        model="microsoft/BioGPT",
+        tokenizer="microsoft/BioGPT",
+        max_new_tokens=100,      # Reduced
+        max_length=400,          # Reduced
+        temperature=0.2,
+        device_map="cpu",        # Force CPU if GPU memory is limited
         return_full_text=False,
         truncation=True,
         do_sample=True,
-        torch_dtype="float32",
+        pad_token_id=1,
+        eos_token_id=2,
         )
     
     # Wrap it in LangChain