File size: 5,985 Bytes
b37a5cd f37c2ef b37a5cd f37c2ef b37a5cd 8439d1b b37a5cd f37c2ef b37a5cd 27f88e4 b37a5cd 27f88e4 b37a5cd c522abc b37a5cd 27f88e4 b37a5cd c522abc b37a5cd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 |
import os
import logging
from typing import Any, List, Mapping, Optional
from gradio_client import Client
from langchain.schema import Document
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import FAISS
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
from langchain.callbacks.manager import CallbackManagerForLLMRun
from langchain.llms.base import LLM
from langchain.chains import RetrievalQA
import streamlit as st
models = '''| Model | Llama2 | Llama2-hf | Llama2-chat | Llama2-chat-hf |
|---|---|---|---|---|
| 70B | [Link](https://huggingface.co/meta-llama/Llama-2-70b) | [Link](https://huggingface.co/meta-llama/Llama-2-70b-hf) | [Link](https://huggingface.co/meta-llama/Llama-2-70b-chat) | [Link](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) |
---'''
DESCRIPTION = """
Welcome to the **YouTube Video Chatbot** powered by the state-of-the-art Llama-2-70b model. Here's what you can do:
- **Transcribe & Understand**: Provide any YouTube video URL, and our system will transcribe it. Our advanced NLP model will then understand the content, ready to answer your questions.
- **Ask Anything**: Based on the video's content, ask any question, and get instant, context-aware answers.
- **Deep Dive**: Our model doesn't just provide generic answers. It understands the context, nuances, and details from the video.
- **Safe & Private**: We value your privacy. The videos you provide are only used for transcription and are not stored or used for any other purpose.
To get started, simply paste a YouTube video URL in the sidebar and start chatting with the model about the video's content. Enjoy the experience!
"""
st.markdown(DESCRIPTION)
def transcribe_video(youtube_url: str, path: str) -> List[Document]:
"""
Transcribe a video and return its content as a Document.
"""
logging.info(f"Transcribing video: {youtube_url}")
client = Client("https://sanchit-gandhi-whisper-jax.hf.space/")
result = client.predict(youtube_url, "translate", True, fn_index=7)
return [Document(page_content=result[1], metadata=dict(page=1))]
def predict(message: str, system_prompt: str = '', temperature: float = 0.7, max_new_tokens: int = 4096,
topp: float = 0.5, repetition_penalty: float = 1.2) -> Any:
"""
Predict a response using a client.
"""
client = Client("https://ysharma-explore-llamav2-with-tgi.hf.space/")
response = client.predict(
message,
system_prompt,
temperature,
max_new_tokens,
topp,
repetition_penalty,
api_name="/chat_1"
)
return response
class LlamaLLM(LLM):
"""
Custom LLM class.
"""
@property
def _llm_type(self) -> str:
return "custom"
def _call(self, prompt: str, stop: Optional[List[str]] = None,
run_manager: Optional[CallbackManagerForLLMRun] = None) -> str:
response = predict(prompt)
return response
@property
def _identifying_params(self) -> Mapping[str, Any]:
"""Get the identifying parameters."""
return {}
PATH = os.path.join(os.path.expanduser("~"), "Data")
def initialize_session_state():
if "youtube_url" not in st.session_state:
st.session_state.youtube_url = ""
if "setup_done" not in st.session_state: # Initialize the setup_done flag
st.session_state.setup_done = False
if "doneYoutubeurl" not in st.session_state:
st.session_state.doneYoutubeurl = ""
def sidebar():
with st.sidebar:
st.markdown(
"## How to use\n"
"1. Enter the YouTube Video URL below🔗\n"
)
st.session_state.youtube_url = st.text_input("YouTube Video URL:")
st.set_page_config(page_title="YouTube Video Chatbot",
layout="centered",
initial_sidebar_state="expanded")
st.title("YouTube Video Chatbot")
sidebar()
initialize_session_state()
# Check if a new YouTube URL is provided
if st.session_state.youtube_url != st.session_state.doneYoutubeurl:
st.session_state.setup_done = False
if st.session_state.youtube_url and not st.session_state.setup_done:
with st.status("Transcribing video..."):
data = transcribe_video(st.session_state.youtube_url, PATH)
with st.status("Running Embeddings..."):
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
docs = text_splitter.split_documents(data)
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-l6-v2")
docsearch = FAISS.from_documents(docs, embeddings)
retriever = docsearch.as_retriever()
retriever.search_kwargs['distance_metric'] = 'cos'
retriever.search_kwargs['k'] = 4
with st.status("Running RetrievalQA..."):
llama_instance = LlamaLLM()
st.session_state.qa = RetrievalQA.from_chain_type(llm=llama_instance, chain_type="stuff", retriever=retriever)
st.session_state.doneYoutubeurl = st.session_state.youtube_url
st.session_state.doneYoutubeurl = st.session_state.youtube_url
st.session_state.setup_done = True # Mark the setup as done for this URL
if "messages" not in st.session_state:
st.session_state.messages = []
for message in st.session_state.messages:
with st.chat_message(message["role"], avatar=("🧑💻" if message["role"] == 'human' else '🦙')):
st.markdown(message["content"])
textinput = st.chat_input("Ask LLama-2-70b anything about the video...")
if prompt := textinput:
st.chat_message("human",avatar = "🧑💻").markdown(prompt)
st.session_state.messages.append({"role": "human", "content": prompt})
with st.status("Requesting Client..."):
response = st.session_state.qa.run(prompt)
with st.chat_message("assistant", avatar='🦙'):
st.markdown(response)
# Add assistant response to chat history
st.session_state.messages.append({"role": "assistant", "content": response}) |