tanuki8x8bchat / app.py
kanhatakeyama's picture
Update app.py
ed2a67d verified
import gradio as gr
from openai import OpenAI
import os
import json
from datetime import datetime
from zoneinfo import ZoneInfo
import uuid
from pathlib import Path
from huggingface_hub import CommitScheduler
openai_api_key = os.getenv('api_key')
openai_api_base = os.getenv('url')
model_name = "weblab-GENIAC/Tanuki-8x8B-dpo-v1.0"
"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
# client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
client = OpenAI(
api_key=openai_api_key,
base_url=openai_api_base,
)
# Define the file where to save the data. Use UUID to make sure not to overwrite existing data from a previous run.
feedback_file = Path("user_feedback/") / f"data_{uuid.uuid4()}.json"
feedback_folder = feedback_file.parent
# Schedule regular uploads. Remote repo and local folder are created if they don't already exist.
scheduler = CommitScheduler(
repo_id="team-hatakeyama-phase2/8x8b-server-original-data4", # Replace with your actual repo ID
repo_type="dataset",
folder_path=feedback_folder,
path_in_repo="data",
every=60, # Upload every 1 minutes
)
def save_or_update_conversation(conversation_id, history,
message, response, message_index, liked=None):
"""
Save or update conversation data in a JSON Lines file.
If the entry already exists (same id and message_index), update the 'label' field.
Otherwise, append a new entry.
"""
with scheduler.lock:
# Read existing data
data = []
if feedback_file.exists():
with feedback_file.open("r") as f:
data = [json.loads(line) for line in f if line.strip()]
# Find if an entry with the same id and message_index exists
#entry_index = next((i for i, entry in enumerate(data) if entry['id'] == conversation_id and entry['message_index'] == message_index), None)
#if entry_index is not None:
## # Update existing entry
# data[entry_index]['label'] = liked
#else:
#always append
if True:
# Append new entry
data.append({
#"id": conversation_id,
"timestamp": datetime.now(ZoneInfo("Asia/Tokyo")).isoformat(),
"history":json.dumps(history,ensure_ascii=False),
"prompt": str(message),
"completion": str(response),
#"message_index": message_index,
"label": liked
})
# Write updated data back to file
with feedback_file.open("w") as f:
for entry in data:
f.write(json.dumps(entry,ensure_ascii=False) + "\n")
def respond(
message,
history: list[tuple[str, str]],
conversation_id,
system_message,
max_tokens,
temperature,
top_p,
):
messages = [
{"role": "system", "content": system_message}]
for val in history:
if val[0]:
messages.append({"role": "user", "content": val[0]})
if val[1]:
messages.append({"role": "assistant", "content": val[1]})
messages.append({"role": "user", "content": message})
response = ""
for chunk in client.chat.completions.create(
model=model_name,
messages=messages,
max_tokens=max_tokens,
stream=True,
temperature=temperature,
top_p=top_p,
stop="### 指示:",
):
token=chunk.choices[0].delta.content
if token is not None:
if response.find("### 指示:")>=0 or token.find("### 指示:")>=0:
response=response.split("### 指示:")[0]
token=token.split("### 指示:")[0]
response=response.replace("### 指示:","")
token=token.replace("### 指示:","")
break
response += token
#response=response.replace("### 指示:","")
yield response
# Save conversation after the full response is generated
message_index = len(history)
save_or_update_conversation(conversation_id,messages, message, response, message_index)
def vote(data: gr.LikeData, history, conversation_id):
"""
Update user feedback (like/dislike) in the local file.
"""
message_index = data.index[0]
liked = data.liked
save_or_update_conversation(conversation_id, history,None, None, message_index, liked)
def create_conversation_id():
return str(uuid.uuid4())
"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
description = """
### [Tanuki-8x8B-dpo-v1.0](https://huggingface.co/weblab-GENIAC/Tanuki-8x8B-dpo-v1.0)との会話(期間限定での公開)
- 9/16停止中
- 人工知能開発のため、原則として**このChatBotの入出力データは全て著作権フリー(CC0)で公開する**ため、ご注意ください。著作物、個人情報、機密情報、誹謗中傷などのデータを入力しないでください。
- **上記の条件に同意する場合のみ**、以下のChatbotを利用してください。
"""
HEADER = description
FOOTER = """### 注意
- コンテクスト長が4096までなので、あまり会話が長くなると、エラーで停止します。ページを再読み込みしてください。
- v1.10"""
def run():
conversation_id = gr.State(create_conversation_id)
chatbot = gr.Chatbot(
elem_id="chatbot",
scale=1,
show_copy_button=True,
height="70%",
layout="panel",
)
with gr.Blocks(fill_height=True) as demo:
gr.Markdown(HEADER)
chat_interface = gr.ChatInterface(
fn=respond,
stop_btn="Stop Generation",
cache_examples=False,
multimodal=False,
chatbot=chatbot,
additional_inputs_accordion=gr.Accordion(
label="Parameters", open=False, render=False
),
additional_inputs=[
conversation_id,
gr.Textbox(value="以下は、タスクを説明する指示です。要求を適切に満たす応答を書きなさい。",
label="System message(試験用: 変えると性能が低下する可能性があります。)",
render=False,),
gr.Slider(
minimum=1,
maximum=4096,
step=1,
value=1024,
label="Max tokens",
visible=True,
render=False,
),
gr.Slider(
minimum=0,
maximum=1,
step=0.1,
value=0.3,
label="Temperature",
visible=True,
render=False,
),
gr.Slider(
minimum=0,
maximum=1,
step=0.1,
value=1.0,
label="Top-p",
visible=True,
render=False,
),
],
analytics_enabled=False,
)
chatbot.like(vote, [chatbot, conversation_id], None)
gr.Markdown(FOOTER)
demo.queue(max_size=256, api_open=True)
demo.launch(share=True, quiet=True)
if __name__ == "__main__":
run()