Spaces:
Running
on
Zero
Running
on
Zero
import torch | |
import gradio as gr | |
import spaces | |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer | |
import os | |
from threading import Thread | |
import random | |
from datasets import load_dataset | |
HF_TOKEN = os.environ.get("HF_TOKEN", None) | |
MODEL_ID = "CohereForAI/c4ai-command-r7b-12-2024" | |
MODELS = os.environ.get("MODELS") | |
MODEL_NAME = MODEL_ID.split("/")[-1] | |
TITLE = "<h1><center>์๋ก์ด ์ผ๋ณธ์ด LLM ๋ชจ๋ธ ์น UI</center></h1>" | |
DESCRIPTION = f""" | |
<h3>๋ชจ๋ธ: <a href="https://huggingface.co/CohereForAI/c4ai-command-r7b-12-2024">CohereForAI/c4ai-command-r7b-12-2024</a></h3> | |
<center> | |
<p> | |
<br> | |
cc-by-nc | |
</p> | |
</center> | |
""" | |
CSS = """ | |
.duplicate-button { | |
margin: auto !important; | |
color: white !important; | |
background: black !important; | |
border-radius: 100vh !important; | |
} | |
h3 { | |
text-align: center; | |
} | |
.chatbox .messages .message.user { | |
background-color: #e1f5fe; | |
} | |
.chatbox .messages .message.bot { | |
background-color: #eeeeee; | |
} | |
""" | |
# ๋ชจ๋ธ๊ณผ ํ ํฌ๋์ด์ ๋ก๋ | |
model = AutoModelForCausalLM.from_pretrained( | |
MODEL_ID, | |
torch_dtype=torch.bfloat16, | |
device_map="auto", | |
) | |
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID) | |
# ๋ฐ์ดํฐ์ ๋ก๋ | |
dataset = load_dataset("elyza/ELYZA-tasks-100") | |
print(dataset) | |
split_name = "train" if "train" in dataset else "test" | |
examples_list = list(dataset[split_name]) | |
examples = random.sample(examples_list, 50) | |
example_inputs = [[example['input']] for example in examples] | |
def stream_chat(message: str, history: list, temperature: float, max_new_tokens: int, top_p: float, top_k: int, penalty: float): | |
print(f'message is - {message}') | |
print(f'history is - {history}') | |
conversation = [] | |
for prompt, answer in history: | |
conversation.extend([{"role": "user", "content": prompt}, {"role": "assistant", "content": answer}]) | |
conversation.append({"role": "user", "content": message}) | |
input_ids = tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True) | |
inputs = tokenizer(input_ids, return_tensors="pt").to(0) | |
streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True) | |
generate_kwargs = dict( | |
inputs, | |
streamer=streamer, | |
top_k=top_k, | |
top_p=top_p, | |
repetition_penalty=penalty, | |
max_new_tokens=max_new_tokens, | |
do_sample=True, | |
temperature=temperature, | |
eos_token_id=[255001], | |
) | |
thread = Thread(target=model.generate, kwargs=generate_kwargs) | |
thread.start() | |
buffer = "" | |
for new_text in streamer: | |
buffer += new_text | |
yield buffer | |
chatbot = gr.Chatbot(height=500) | |
with gr.Blocks(css=CSS) as demo: | |
gr.HTML(TITLE) | |
gr.HTML(DESCRIPTION) | |
gr.ChatInterface( | |
fn=stream_chat, | |
chatbot=chatbot, | |
fill_height=True, | |
theme="soft", | |
additional_inputs_accordion=gr.Accordion(label="โ๏ธ ๋งค๊ฐ๋ณ์", open=False, render=False), | |
additional_inputs=[ | |
gr.Slider( | |
minimum=0, | |
maximum=1, | |
step=0.1, | |
value=0.8, | |
label="์จ๋", | |
render=False, | |
), | |
gr.Slider( | |
minimum=128, | |
maximum=1000000, | |
step=1, | |
value=100000, | |
label="์ต๋ ํ ํฐ ์", | |
render=False, | |
), | |
gr.Slider( | |
minimum=0.0, | |
maximum=1.0, | |
step=0.1, | |
value=0.8, | |
label="์์ ํ๋ฅ ", | |
render=False, | |
), | |
gr.Slider( | |
minimum=1, | |
maximum=20, | |
step=1, | |
value=20, | |
label="์์ K", | |
render=False, | |
), | |
gr.Slider( | |
minimum=0.0, | |
maximum=2.0, | |
step=0.1, | |
value=1.0, | |
label="๋ฐ๋ณต ํจ๋ํฐ", | |
render=False, | |
), | |
], | |
examples=[ | |
["์์ด์ ์ฌ๋ฆ๋ฐฉํ ๊ณผํ ํ๋ก์ ํธ๋ฅผ ์ํ 5๊ฐ์ง ์์ด๋์ด๋ฅผ ์ฃผ์ธ์."], | |
["๋งํฌ๋ค์ด์ ์ฌ์ฉํ์ฌ ๋ธ๋ ์ดํฌ์์ ๊ฒ์ ๋ง๋ค๊ธฐ ํํ ๋ฆฌ์ผ์ ์์ฑํด์ฃผ์ธ์."], | |
["์ด๋ฅ๋ ฅ์ ๊ฐ์ง ์ฃผ์ธ๊ณต์ SF ์ด์ผ๊ธฐ ์๋๋ฆฌ์ค๋ฅผ ์์ฑํด์ฃผ์ธ์. ๋ณต์ ์ค์ , ํ ๋ง์ ๋ก๊ทธ๋ผ์ธ์ ๋ ผ๋ฆฌ์ ์ผ๋ก ์ฌ์ฉํด์ฃผ์ธ์"], | |
["์์ด์ ์ฌ๋ฆ๋ฐฉํ ์์ ์ฐ๊ตฌ๋ฅผ ์ํ 5๊ฐ์ง ์์ด๋์ด์ ๊ทธ ๋ฐฉ๋ฒ์ ๊ฐ๋จํ ์๋ ค์ฃผ์ธ์."], | |
["ํผ์ฆ ๊ฒ์ ์คํฌ๋ฆฝํธ ์์ฑ์ ์ํ ์กฐ์ธ ๋ถํ๋๋ฆฝ๋๋ค"], | |
["๋งํฌ๋ค์ด ํ์์ผ๋ก ๋ธ๋ก ๊นจ๊ธฐ ๊ฒ์ ์ ์ ๊ต๊ณผ์๋ฅผ ์์ฑํด์ฃผ์ธ์"], | |
["์ค๋ฒ ๅทๆณ๋ฅผ ์๊ฐํด์ฃผ์ธ์"], | |
["์ผ๋ณธ์ด ๊ด์ฉ๊ตฌ, ์๋ด์ ๊ดํ ์ํ ๋ฌธ์ ๋ฅผ ๋ง๋ค์ด์ฃผ์ธ์"], | |
["๋๋ผ์๋ชฝ์ ๋ฑ์ฅ์ธ๋ฌผ์ ์๋ ค์ฃผ์ธ์"], | |
["์ค์ฝ๋ ธ๋ฏธ์ผํค ๋ง๋๋ ๋ฐฉ๋ฒ์ ์๋ ค์ฃผ์ธ์"], | |
["๋ฌธ์ 9.11๊ณผ 9.9 ์ค ์ด๋ ๊ฒ์ด ๋ ํฐ๊ฐ์? step by step์ผ๋ก ๋ ผ๋ฆฌ์ ์ผ๋ก ์๊ฐํด์ฃผ์ธ์."], | |
], | |
cache_examples=False, | |
) | |
if __name__ == "__main__": | |
demo.launch() |