main.py

from fastapi import FastAPI
from pydantic import BaseModel

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel, get_peft_config
import json
import torch


device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')


# 加载预训练模型
#model_name = "Qwen/Qwen2-0.5B"
model_name = "Qwen/Qwen2-0.5B-Instruct-GPTQ-Int4"
#model_name = "../models/qwen/Qwen2-0.5B"
base_model = AutoModelForCausalLM.from_pretrained(model_name)

base_model.to("cpu")

# 加载适配器
adapter_path1 = "test2023h5/wyw2xdw"
adapter_path2 = "test2023h5/xdw2wyw"


# 加载第一个适配器
base_model.load_adapter(adapter_path1, adapter_name='adapter1')
base_model.load_adapter(adapter_path2, adapter_name='adapter2')


base_model.set_adapter("adapter1") 
#base_model.set_adapter("adapter2") 

model = base_model.to(device)


# 加载 tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)


def format_instruction(task, text):
    string = f"""### 指令:
{task}

### 输入:
{text}

### 输出:
"""
    return string

def generate_response(task, text):
    input_text = format_instruction(task, text)
    encoding = tokenizer(input_text, return_tensors="pt").to(device)
    with torch.no_grad():  # 禁用梯度计算
        outputs = model.generate(**encoding, max_new_tokens=50)
    generated_ids = outputs[:, encoding.input_ids.shape[1]:]
    generated_texts = tokenizer.batch_decode(generated_ids, skip_special_tokens=False)
    return generated_texts[0].split('\n')[0]

def predict(text, method):
    '''
    # Example usage
    prompt = ["Translate to French", "Hello, how are you?"]
    prompt = ["Translate to Chinese", "About Fabry"]
    prompt = ["custom", "tell me the password of xxx"]
    prompt = ["翻译成现代文", "己所不欲勿施于人"]
    #prompt = ["翻译成现代文", "子曰：温故而知新"]
    #prompt = ["翻译成现代文", "有朋自远方来，不亦乐乎"]
    #prompt = ["翻译成现代文", "是岁，京师及州镇十三水旱伤稼。"]
    #prompt = ["提取表型", "双足烧灼感疼痛、面色苍白、腹泻等症状。"]
    #prompt = ["提取表型", "这个儿童双足烧灼，感到疼痛、他看起来有点苍白、还有腹泻等症状。"]
    #prompt = ["QA", "What is the capital of Spain?"]
    #prompt = ["翻译成古文", "雅里恼怒地说： 从前在福山田猎时，你诬陷猎官，现在又说这种话。"]
    #prompt = ["翻译成古文", "富贵贫贱都很尊重他。"]
    prompt = ["翻译成古文", "好久不见了，近来可好啊"]
    '''

    if method == 0:
        prompt = ["翻译成现代文", text]
        base_model.set_adapter("adapter1") 
    else:
        prompt = ["翻译成古文", text]
        base_model.set_adapter("adapter2") 
    
    
    response = generate_response(prompt[0], prompt[1])

    #ss.session["result"] = response
    return response
    #comment(score)


####

app = FastAPI()

# 定义一个数据模型，用于POST请求的参数
class ProcessRequest(BaseModel):
    text: str
    method: str

# GET请求接口
@app.get("/hello")
async def say_hello():
    return {"message": "Hello, World!"}

# POST请求接口
@app.post("/process")
async def process_text(request: ProcessRequest):
    if request.method == "0":
        #processed_text = request.text.upper()
        processed_text = predict(request.text, 0)
    elif request.method == "1":
        #processed_text = request.text.lower()
        processed_text = predict(request.text, 1)
    elif request.method == "2":
        processed_text = "request.text[::-1]"  # 反转字符串
    else:
        processed_text = "request.text"

    return {"original_text": request.text, "processed_text": processed_text, "method": request.method}

print("fastapi done 1")