Update app.py

app.py

@@ -1,7 +1,8 @@
+from transformers import GPT2LMHeadModel, GPT2Tokenizer
 from langchain import PromptTemplate
 from langchain.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores import FAISS
-from langchain_community.llms import CTransformers
+from langchain_community.llms import CTransformers  # You might need to change this if GPT-2 isn't directly supported
 from langchain.chains import RetrievalQA
 import gradio as gr
 from huggingface_hub import hf_hub_download
@@ -10,17 +11,18 @@ DB_FAISS_PATH = "vectorstores/db_faiss"
 
 def load_llm():
     """
-    Load the LLaMA model for the language model.
+    Load the GPT-2 model for the language model.
     """
-    model_name = 'TheBloke/Llama-2-7B-Chat-GGML'
-    model_path = hf_hub_download(repo_id=model_name, filename='llama-2-7b-chat.ggmlv3.q8_0.bin', cache_dir='./models')
-    llm = CTransformers(
-        model=model_path,
-        model_type="llama",
-        max_new_tokens=512,
-        temperature=0.5
-    )
-    return llm
+    try:
+        print("Downloading or loading the GPT-2 model and tokenizer...")
+        model_name = 'gpt2'
+        model = GPT2LMHeadModel.from_pretrained(model_name)
+        tokenizer = GPT2Tokenizer.from_pretrained(model_name)
+        print("Model and tokenizer successfully loaded!")
+        return model, tokenizer
+    except Exception as e:
+        print(f"An error occurred while loading the model: {e}")
+        return None, None
 
 def set_custom_prompt():
     """
@@ -38,12 +40,19 @@ Helpful answer:
     prompt = PromptTemplate(template=custom_prompt_template, input_variables=['context', 'question'])
     return prompt
 
-def retrieval_QA_chain(llm, prompt, db):
+def retrieval_QA_chain(llm, tokenizer, prompt, db):
     """
     Create a RetrievalQA chain with the specified LLM, prompt, and vector store.
     """
+    def generate_answer(query):
+        # Tokenize the input query
+        inputs = tokenizer.encode(query, return_tensors='pt')
+        # Generate response
+        outputs = llm.generate(inputs, max_length=512, temperature=0.5)
+        return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
     qachain = RetrievalQA.from_chain_type(
-        llm=llm,
+        llm=generate_answer,
         chain_type="stuff",
         retriever=db.as_retriever(search_kwargs={'k': 2}),
         return_source_documents=True,
@@ -57,9 +66,12 @@ def qa_bot():
     """
     embeddings = HuggingFaceEmbeddings(model_name='sentence-transformers/all-miniLM-L6-V2', model_kwargs={'device': 'cpu'})
     db = FAISS.load_local(DB_FAISS_PATH, embeddings, allow_dangerous_deserialization=True)
-    llm = load_llm()
+    model, tokenizer = load_llm()
     qa_prompt = set_custom_prompt()
-    qa = retrieval_QA_chain(llm, qa_prompt, db)
+    if model and tokenizer:
+        qa = retrieval_QA_chain(model, tokenizer, qa_prompt, db)
+    else:
+        qa = None
     return qa
 
 bot = qa_bot()
@@ -69,14 +81,17 @@ def chatbot_response(message, history):
     Generate a response from the chatbot based on the user input and conversation history.
     """
     try:
-        response = bot({'query': message})
-        answer = response["result"]
-        sources = response["source_documents"]
-        if sources:
-            answer += f"\nSources: {sources}"
+        if bot:
+            response = bot({'query': message})
+            answer = response["result"]
+            sources = response.get("source_documents", [])
+            if sources:
+                answer += f"\nSources: {sources}"
+            else:
+                answer += "\nNo sources found"
+            history.append((message, answer))
         else:
-            answer += "\nNo sources found"
-        history.append((message, answer))
+            history.append((message, "Model is not loaded properly."))
     except Exception as e:
         history.append((message, f"An error occurred: {str(e)}"))
     return history, history
@@ -97,4 +112,4 @@ demo = gr.Interface(
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()