order imports, remove special case models and remove comments

app.py

@@ -1,29 +1,23 @@
-import gradio as gr
 import os
-
-from langchain_community.document_loaders import PyPDFLoader
-from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain_community.vectorstores import Chroma
-from langchain.chains import ConversationalRetrievalChain
-from langchain_community.embeddings import HuggingFaceEmbeddings
-from langchain_community.llms import HuggingFacePipeline
-from langchain.chains import ConversationChain
-from langchain.memory import ConversationBufferMemory
-from langchain_community.llms import HuggingFaceEndpoint
-
+import re
 from pathlib import Path
-import chromadb
-from unidecode import unidecode
 
-from transformers import AutoTokenizer
-import transformers
+import accelerate
+import chromadb
+import gradio as gr
 import torch
 import tqdm
-import accelerate
-import re
-
+import transformers
+from langchain.chains import ConversationalRetrievalChain, ConversationChain
+from langchain.memory import ConversationBufferMemory
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_community.llms import HuggingFaceEndpoint, HuggingFacePipeline
+from langchain_community.vectorstores import Chroma
+from transformers import AutoTokenizer
+from unidecode import unidecode
 
-# default_persist_directory = './chroma_HF/'
 list_llm = [
     "mistralai/Mistral-7B-Instruct-v0.2",
     "mistralai/Mixtral-8x7B-Instruct-v0.1",
@@ -31,8 +25,6 @@ list_llm = [
     "google/gemma-7b-it",
     "google/gemma-2b-it",
     "HuggingFaceH4/zephyr-7b-beta",
-    "microsoft/phi-2",
-    "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
     "tiiuae/falcon-7b-instruct",
     "google/flan-t5-xxl",
 ]
@@ -65,7 +57,6 @@ def create_db(splits, collection_name):
         embedding=embedding,
         client=new_client,
         collection_name=collection_name,
-        # persist_directory=default_persist_directory
     )
     return vectordb
 
@@ -73,10 +64,7 @@ def create_db(splits, collection_name):
 # Load vector database
 def load_db():
     embedding = HuggingFaceEmbeddings()
-    vectordb = Chroma(
-        # persist_directory=default_persist_directory,
-        embedding_function=embedding
-    )
+    vectordb = Chroma(embedding_function=embedding)
     return vectordb
 
 
@@ -85,64 +73,20 @@ def initialize_llmchain(
     llm_model, temperature, max_tokens, top_k, vector_db, progress=gr.Progress()
 ):
     progress(0.1, desc="Initializing HF tokenizer...")
-    # HuggingFacePipeline uses local model
-    # Note: it will download model locally...
-    # tokenizer=AutoTokenizer.from_pretrained(llm_model)
-    # progress(0.5, desc="Initializing HF pipeline...")
-    # pipeline=transformers.pipeline(
-    #     "text-generation",
-    #     model=llm_model,
-    #     tokenizer=tokenizer,
-    #     torch_dtype=torch.bfloat16,
-    #     trust_remote_code=True,
-    #     device_map="auto",
-    #     # max_length=1024,
-    #     max_new_tokens=max_tokens,
-    #     do_sample=True,
-    #     top_k=top_k,
-    #     num_return_sequences=1,
-    #     eos_token_id=tokenizer.eos_token_id
-    #     )
-    # llm = HuggingFacePipeline(pipeline=pipeline, model_kwargs={'temperature': temperature})
 
     # HuggingFaceHub uses HF inference endpoints
     progress(0.5, desc="Initializing HF Hub...")
-    # Use of trust_remote_code as model_kwargs
-    # Warning: langchain issue
-    # URL: https://github.com/langchain-ai/langchain/issues/6080
     if llm_model == "mistralai/Mixtral-8x7B-Instruct-v0.1":
         llm = HuggingFaceEndpoint(
             repo_id=llm_model,
-            # model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens, "top_k": top_k, "load_in_8bit": True}
             temperature=temperature,
             max_new_tokens=max_tokens,
             top_k=top_k,
             load_in_8bit=True,
         )
-    elif llm_model == "microsoft/phi-2":
-        # raise gr.Error("phi-2 model requires 'trust_remote_code=True', currently not supported by langchain HuggingFaceHub...")
-        llm = HuggingFaceEndpoint(
-            repo_id=llm_model,
-            # model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens, "top_k": top_k, "trust_remote_code": True, "torch_dtype": "auto"}
-            temperature=temperature,
-            max_new_tokens=max_tokens,
-            top_k=top_k,
-            trust_remote_code=True,
-            torch_dtype="auto",
-        )
-    elif llm_model == "TinyLlama/TinyLlama-1.1B-Chat-v1.0":
-        llm = HuggingFaceEndpoint(
-            repo_id=llm_model,
-            # model_kwargs={"temperature": temperature, "max_new_tokens": 250, "top_k": top_k}
-            temperature=temperature,
-            max_new_tokens=250,
-            top_k=top_k,
-        )
     else:
         llm = HuggingFaceEndpoint(
             repo_id=llm_model,
-            # model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens, "top_k": top_k, "trust_remote_code": True, "torch_dtype": "auto"}
-            # model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens, "top_k": top_k}
             temperature=temperature,
             max_new_tokens=max_tokens,
             top_k=top_k,
@@ -154,15 +98,14 @@ def initialize_llmchain(
     )
     # retriever=vector_db.as_retriever(search_type="similarity", search_kwargs={'k': 3})
     retriever = vector_db.as_retriever()
+
     progress(0.8, desc="Defining retrieval chain...")
     qa_chain = ConversationalRetrievalChain.from_llm(
         llm,
         retriever=retriever,
         chain_type="stuff",
         memory=memory,
-        # combine_docs_chain_kwargs={"prompt": your_prompt})
         return_source_documents=True,
-        # return_generated_question=False,
         verbose=False,
     )
     progress(0.9, desc="Done!")
@@ -197,22 +140,19 @@ def create_collection_name(filepath):
     return collection_name
 
 
-# Initialize database
 def initialize_database(
     list_file_obj, chunk_size, chunk_overlap, progress=gr.Progress()
 ):
-    # Create list of documents (when valid)
     list_file_path = [x.name for x in list_file_obj if x is not None]
-    # Create collection_name for vector database
     progress(0.1, desc="Creating collection name...")
     collection_name = create_collection_name(list_file_path[0])
+
     progress(0.25, desc="Loading document...")
-    # Load document and create splits
     doc_splits = load_doc(list_file_path, chunk_size, chunk_overlap)
-    # Create or load vector database
+
     progress(0.5, desc="Generating vector database...")
-    # global vector_db
     vector_db = create_db(doc_splits, collection_name)
+
     progress(0.9, desc="Done!")
     return vector_db, collection_name, "Complete!"
 
@@ -220,7 +160,6 @@ def initialize_database(
 def initialize_LLM(
     llm_option, llm_temperature, max_tokens, top_k, vector_db, progress=gr.Progress()
 ):
-    # print("llm_option",llm_option)
     llm_name = list_llm[llm_option]
     print("llm_name: ", llm_name)
     qa_chain = initialize_llmchain(
@@ -239,7 +178,6 @@ def format_chat_history(message, chat_history):
 
 def conversation(qa_chain, message, history):
     formatted_chat_history = format_chat_history(message, history)
-    # print("formatted_chat_history",formatted_chat_history)
 
     # Generate response using QA chain
     response = qa_chain({"question": message, "chat_history": formatted_chat_history})
@@ -247,15 +185,13 @@ def conversation(qa_chain, message, history):
     if response_answer.find("Helpful Answer:") != -1:
         response_answer = response_answer.split("Helpful Answer:")[-1]
     response_sources = response["source_documents"]
+    # Langchain sources are zero-based
     response_source1 = response_sources[0].page_content.strip()
     response_source2 = response_sources[1].page_content.strip()
     response_source3 = response_sources[2].page_content.strip()
-    # Langchain sources are zero-based
     response_source1_page = response_sources[0].metadata["page"] + 1
     response_source2_page = response_sources[1].metadata["page"] + 1
     response_source3_page = response_sources[2].metadata["page"] + 1
-    # print ('chat response: ', response_answer)
-    # print('DB source', response_sources)
 
     # Append user message and response to chat history
     new_history = history + [(message, response_answer)]
@@ -278,8 +214,6 @@ def upload_file(file_obj):
     for idx, file in enumerate(file_obj):
         file_path = file_obj.name
         list_file_path.append(file_path)
-    # print(file_path)
-    # initialize_database(file_path, progress)
     return list_file_path