|
import gradio as gr |
|
import torch |
|
from transformers import ( |
|
AutoModelForCausalLM, |
|
AutoTokenizer, |
|
TextIteratorStreamer, |
|
BitsAndBytesConfig, |
|
) |
|
import os |
|
from threading import Thread |
|
import spaces |
|
import time |
|
|
|
token = os.environ["HF_TOKEN"] |
|
|
|
quantization_config = BitsAndBytesConfig( |
|
load_in_4bit=True, bnb_4bit_compute_dtype=torch.float16 |
|
) |
|
|
|
model = AutoModelForCausalLM.from_pretrained( |
|
"NousResearch/Hermes-2-Pro-Llama-3-8B", quantization_config=quantization_config, token=token |
|
) |
|
tok = AutoTokenizer.from_pretrained("NousResearch/Hermes-2-Pro-Llama-3-8B", token=token) |
|
terminators = [ |
|
tok.eos_token_id, |
|
tok.convert_tokens_to_ids("<|eot_id|>") |
|
] |
|
|
|
if torch.cuda.is_available(): |
|
device = torch.device("cuda") |
|
print(f"Using GPU: {torch.cuda.get_device_name(device)}") |
|
else: |
|
device = torch.device("cpu") |
|
print("Using CPU") |
|
|
|
|
|
|
|
|
|
|
|
@spaces.GPU(duration=150) |
|
def chat(message, history, temperature,do_sample, max_tokens): |
|
chat = [] |
|
for item in history: |
|
chat.append({"role": "user", "content": item[0]}) |
|
if item[1] is not None: |
|
chat.append({"role": "assistant", "content": item[1]}) |
|
chat.append({"role": "user", "content": message}) |
|
messages = tok.apply_chat_template(chat, tokenize=False, add_generation_prompt=True) |
|
model_inputs = tok([messages], return_tensors="pt").to(device) |
|
streamer = TextIteratorStreamer( |
|
tok, timeout=10.0, skip_prompt=True, skip_special_tokens=True |
|
) |
|
generate_kwargs = dict( |
|
model_inputs, |
|
streamer=streamer, |
|
max_new_tokens=max_tokens, |
|
do_sample=True, |
|
temperature=temperature, |
|
eos_token_id=terminators, |
|
) |
|
|
|
if temperature == 0: |
|
generate_kwargs['do_sample'] = False |
|
|
|
t = Thread(target=model.generate, kwargs=generate_kwargs) |
|
t.start() |
|
|
|
partial_text = "" |
|
for new_text in streamer: |
|
partial_text += new_text |
|
yield partial_text |
|
|
|
tokens = len(tok.tokenize(partial_text)) |
|
yield partial_text |
|
|
|
|
|
demo = gr.ChatInterface( |
|
fn=chat, |
|
examples=[["Write me a poem about Machine Learning."]], |
|
|
|
additional_inputs_accordion=gr.Accordion( |
|
label="⚙️ Parameters", open=False, render=False |
|
), |
|
additional_inputs=[ |
|
gr.Slider( |
|
minimum=0, maximum=1, step=0.1, value=0.9, label="Temperature", render=False |
|
), |
|
gr.Checkbox(label="Sampling",value=True), |
|
gr.Slider( |
|
minimum=128, |
|
maximum=4096, |
|
step=1, |
|
value=512, |
|
label="Max new tokens", |
|
render=False, |
|
), |
|
], |
|
stop_btn="Stop Generation", |
|
title="Chat With LLMs", |
|
description="Now Running [NousResearch/Hermes-2-Pro-Llama-3-8B](https://huggingface.co/NousResearch/Hermes-2-Pro-Llama-3-8B) in 4bit" |
|
) |
|
demo.launch() |