QiyuWu's picture
Update app.py
9016f6f verified
import os
import gradio as gr
import copy
from llama_cpp import Llama
from huggingface_hub import hf_hub_download
llm = Llama(
model_path = "./dpo-internlm2-1_8b.Q4_K_M.gguf",
n_ctx=2048,
n_threads=2,
seed=57199,
batch_size=1024,
cache=True
)
def generate_text(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
):
temp = ""
input_prompt = f"<s> <|im_start|>system:{system_message}<|im_end|>"
for interaction in history:
input_prompt = input_prompt + "<|im_start|>user:" + str(interaction[0]) + "<|im_end|><|im_start|>" + "assistant:" + str(interaction[1]) + "<|im_end|>"
input_prompt = input_prompt + "<|im_start|>user:" + str(message) + "<|im_end|><|im_start|>assistant:"
output = llm(
input_prompt,
temperature=temperature,
top_p=top_p,
top_k=40,
repeat_penalty=1.2,
max_tokens=max_tokens,
stop=["<|im_end|>", "[UNUSED_TOKEN_145]"],
stream=True,
)
for out in output:
stream = copy.deepcopy(out)
temp += stream["choices"][0]["text"]
yield temp
demo = gr.ChatInterface(
generate_text,
title="DPO-Internlm2-1_8B",
description="Running LLM with llama-cpp-python",
theme = gr.themes.Soft(),
examples=[
['How to setup a human base on Mars? Give short answer.'],
['Explain theory of relativity to me like I’m 8 years old.'],
['What is 9,000 * 9,000?'],
['Write a pun-filled happy birthday message to my friend Alex.'],
['Justify why a penguin might make a good king of the jungle.']
],
cache_examples=False,
retry_btn=None,
undo_btn="Delete Previous",
clear_btn="Clear",
additional_inputs=[
gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
gr.Slider(minimum=0.1, maximum=4.0, value=0.3, step=0.1, label="Temperature"),
gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.90,
step=0.05,
label="Top-p (nucleus sampling)",
),
],
)
if __name__ == "__main__":
demo.launch()