app.py

# coding=utf-8
from typing import Dict
from typing import List
from typing import Tuple
from typing import Union
from pathlib import Path
from src.logger import LoggerFactory
from src.prompt_concat import GetManualTestSamples, CreateTestDataset
from src.utils import decode_csv_to_json, load_json, save_to_json
from threading import Thread
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    GenerationConfig,
    TextIteratorStreamer,
)
from typing import List

import gradio as gr
import logging
import os
import shutil
import torch
import warnings
import random
import spaces

logger = LoggerFactory.create_logger(name="test", level=logging.INFO)
warnings.filterwarnings('ignore', category=UserWarning, message='TypedStorage is deprecated')

MODEL_PATH = os.environ.get('MODEL_PATH', 'IndexTeam/Index-1.9B-Character')
TOKENIZER_PATH = os.environ.get("TOKENIZER_PATH", MODEL_PATH)

tokenizer = AutoTokenizer.from_pretrained(TOKENIZER_PATH, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, torch_dtype=torch.float16, device_map="auto",
                                             trust_remote_code=True)

character_path = "./character"

def _resolve_path(path: Union[str, Path]) -> Path:
    return Path(path).expanduser().resolve()

# logger = LoggerFactory.create_logger(name="test", level=logging.INFO)
# warnings.filterwarnings('ignore', category=UserWarning, message='TypedStorage is deprecated')

# config_data = load_json("config/config.json")
# model_path = config_data["huggingface_local_path"]
# character_path = "./character"
# tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
# model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.float16, device_map="auto",
#                                              trust_remote_code=True)


def generate_with_question(question, role_name, role_file_path):
    question_in = "\n".join(["\n".join(pair) for pair in question])

    g = GetManualTestSamples(
        role_name=role_name,
        role_data_path=f"./character/{role_file_path}.json",
        save_samples_dir="./character",
        save_samples_path= role_file_path + "_rag.json",
        prompt_path="./prompt/dataset_character.txt",
        max_seq_len=4000
    )
    g.get_qa_samples_by_query(
        questions_query=question_in,
        keep_retrieve_results_flag=True
    )


def create_datasets(role_name, role_file_path):
    testset = []
    role_samples_path = os.path.join("./character", role_file_path + "_rag.json")

    c = CreateTestDataset(role_name=role_name,
                          role_samples_path=role_samples_path,
                          role_data_path=role_samples_path,
                          prompt_path="./prompt/dataset_character.txt"
                          )
    res = c.load_samples()
    testset.extend(res)
    save_to_json(testset, f"./character/{role_file_path}_测试问题.json")

@spaces.GPU
def hf_gen(dialog: List, role_name, role_file_path, top_k, top_p, temperature, repetition_penalty, max_dec_len):
    generate_with_question(dialog, role_name,role_file_path)
    create_datasets(role_name,role_file_path)

    json_data = load_json(f"{character_path}/{role_file_path}_测试问题.json")[0]
    text = json_data["input_text"]
    inputs = tokenizer(text, return_tensors="pt")

    if torch.cuda.is_available():
        model.to("cuda")
        inputs.to("cuda")

    streamer = TextIteratorStreamer(tokenizer, **tokenizer.init_kwargs)
    generation_kwargs = dict(
        inputs,
        do_sample=True,
        top_k=int(top_k),
        top_p=float(top_p),
        temperature=float(temperature),
        repetition_penalty=float(repetition_penalty),
        max_new_tokens=int(max_dec_len),
        pad_token_id=tokenizer.eos_token_id,
        streamer=streamer,
    )

    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()
    answer = ""
    for new_text in streamer:
        answer += new_text
        yield answer[len(text):]

@spaces.GPU
def generate(chat_history: List, query, role_name, role_desc, role_file_path, top_k, top_p, temperature, repetition_penalty, max_dec_len):
    """generate after hitting "submit" button

    Args:
        chat_history (List): [[q_1, a_1], [q_2, a_2], ..., [q_n, a_n]]. list that stores all QA records
        query (str): query of current round
        top_p (float): only the smallest set of most probable tokens with probabilities that add up to top_p or higher are kept for generation.
        temperature (float): strictly positive float value used to modulate the logits distribution.
        max_dec_len (int): The maximum numbers of tokens to generate.

    Yields:
        List: [[q_1, a_1], [q_2, a_2], ..., [q_n, a_n], [q_n+1, a_n+1]]. chat_history + QA of current round.
    """
    assert query != "", "Input must not be empty!!!"
    # apply chat template
    chat_history.append([f"user:{query}", ""])
    if role_name == "三三":
        role_file_path = "三三"
    for answer in hf_gen(chat_history, role_name,role_file_path, top_k, top_p, temperature, repetition_penalty, max_dec_len):
        chat_history[-1][1] = role_name + ":" + answer
        yield gr.update(value=""), chat_history

@spaces.GPU
def regenerate(chat_history: List,role_name, role_description, role_file_path, top_k, top_p, temperature, repetition_penalty, max_dec_len):
    """re-generate the answer of last round's query

    Args:
        chat_history (List): [[q_1, a_1], [q_2, a_2], ..., [q_n, a_n]]. list that stores all QA records
        top_p (float): only the smallest set of most probable tokens with probabilities that add up to top_p or higher are kept for generation.
        temperature (float): strictly positive float value used to modulate the logits distribution.
        max_dec_len (int): The maximum numbers of tokens to generate.

    Yields:
        List: [[q_1, a_1], [q_2, a_2], ..., [q_n, a_n]]. chat_history
    """
    assert len(chat_history) >= 1, "History is empty. Nothing to regenerate!!"
    if len(chat_history[-1]) > 1:
        chat_history[-1][1] = ""
    # apply chat template
    if role_name == "三三":
        role_file_path = "三三"
    for answer in hf_gen(chat_history, role_name,role_file_path, top_k, top_p, temperature, repetition_penalty, max_dec_len):
        chat_history[-1][1] = role_name + ":" + answer
        yield gr.update(value=""), chat_history

def clear_history():
    """clear all chat history

    Returns:
        List: empty chat history
    """
    torch.cuda.empty_cache()
    return []

def delete_current_user(user_role_path):
    try:
        role_upload_path = os.path.join(character_path, user_role_path + ".csv")
        role_path = os.path.join(character_path, user_role_path + ".json")
        rag_path = os.path.join(character_path, user_role_path + "_rag.json")
        question_path = os.path.join(character_path, user_role_path + "_测试问题.json")

        files_to_delete = [role_upload_path, role_path, rag_path, question_path]

        for file_path in files_to_delete:
            os.remove(file_path)
    except Exception as e:
        print(e)

# launch gradio demo
with gr.Blocks(theme="soft") as demo:
    gr.Markdown("""# Index-1.9B RolePlay Gradio Demo""")

    with gr.Row():
        with gr.Column(scale=1):
            top_k = gr.Slider(0, 10, value=5, step=1, label="top_k")
            top_p = gr.Slider(0, 1, value=0.8, step=0.8, label="top_p")
            temperature = gr.Slider(0.1, 2.0, value=0.85, step=0.1, label="temperature")
            repetition_penalty = gr.Slider(0.1, 2.0, value=1.0, step=0.1, label="repetition_penalty")
            max_dec_len = gr.Slider(1, 4096, value=512, step=1, label="max_dec_len")
            file_input = gr.File(label="上传角色对话语料(.csv)")
            role_description = gr.Textbox(label="您创建的角色描述", placeholder="输入角色描述", lines=2)
            upload_button = gr.Button("生成角色!")


            new_path = gr.State()
            def generate_file(file_obj, role_info):
                random.seed()
                alphabet = 'abcdefghijklmnopqrstuvwxyz!@#$%^&*()'
                random_char = "".join(random.choice(alphabet) for _ in range(10))

                role_name = os.path.basename(file_obj).split(".")[0]
                new_path = role_name + random_char
                new_save_path = os.path.join(character_path, new_path+".csv")
                shutil.copy(file_obj, new_save_path)
                 
                new_file_path = os.path.join(character_path, new_path)
                decode_csv_to_json(os.path.join(character_path, new_path + ".csv"), role_name, role_info,
                                new_file_path + ".json" )
                gr.Info(f"{role_name}生成成功")
                return new_path
            
            upload_button.click(generate_file, inputs=[file_input, role_description],outputs=new_path)
        with gr.Column(scale=10):
            chatbot = gr.Chatbot(bubble_full_width=False, height=400, label='Index-1.9B')
            with gr.Row():
                role_name = gr.Textbox(label="对话的角色名字", value="三三", placeholder="如果您没有创建角色，可以直接输入三三。如果已经创建好了对应的角色，请在这里输入角色的名称!", lines=2)
            user_input = gr.Textbox(label="用户问题", placeholder="输入你的问题!", lines=2)
            with gr.Row():
                submit = gr.Button("🚀 Submit")
                clear = gr.Button("🧹 Clear")
                regen = gr.Button("🔄 Regenerate")

    submit.click(generate, inputs=[chatbot, user_input, role_name, role_description, new_path, top_k, top_p, temperature,
                                    repetition_penalty, max_dec_len],
                    outputs=[user_input, chatbot])
    regen.click(regenerate,
                inputs=[chatbot, role_name, role_description, new_path, top_k, top_p, temperature, repetition_penalty,
                        max_dec_len],
                outputs=[user_input, chatbot])
    clear.click(clear_history, inputs=[], outputs=[chatbot])

demo.queue().launch()