Update app.py

app.py

@@ -141,21 +141,19 @@ def user_input(user_question, api_key):
     chain = get_conversational_chain()
     response_gemini = chain({"input_documents": docs, "question": user_question}, return_only_outputs=True)
     
-    # Initialize the Hugging Face text generation pipeline with your custom model
-    pipeline = TextGeneratorPipeline(model="Writer/palmyra-small")
+    # Initialize the Hugging Face model and tokenizer
+    model_name_or_path = "Writer/palmyra-small"
+    model = AutoModelForCausalLM.from_pretrained(model_name_or_path)
+    tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
 
-    # Prompt template for making the response more conversational
-    prompt_template = f"""
-    Transform the following response into a more conversational tone without adding new information:
-    
-    Response:
-    {response_gemini["output_text"]}
-    
-    Transformed Response:
-    """
+    # Tokenize the prompt
+    inputs = tokenizer(prompt_template.format(response_gemini["output_text"]), return_tensors="pt", max_length=100, truncation=True)
 
     # Generate the transformed response using the Hugging Face model
-    transformed_response = pipeline(prompt=prompt_template, max_length=100)
+    outputs = model.generate(**inputs)
+
+    # Decode the generated response
+    transformed_response = tokenizer.decode(outputs[0], skip_special_tokens=True)
 
     # Display the transformed response
     st.write("Reply: ", transformed_response)
@@ -166,6 +164,7 @@ def user_input(user_question, api_key):
 
 
 
+
 def main():
     st.header("RAG based LLM Application")