Codellama Chat 13b Chinese

🚀 CodeLlaMa模型的中文化版本 （支持多輪對話）

本項目是CodeLlaMa模型的中文化版本，支持多輪對話。CodeLlaMa專門用於代碼助手，與ChineseLlaMa不同，它適用於代碼類問題的回覆。

🚀 快速開始

關聯倉庫

關聯Github倉庫：CodeLLaMA-chat

推理代碼

用於多輪對話的推理代碼如下，你可以直接複製運行，默認會自動拉取該模型權重：

# from Firefly
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch


def main():
    model_name = 'shareAI/CodeLLaMA-chat-13b-Chinese'

    device = 'cuda'
    max_new_tokens = 500    # 每輪對話最多生成多少個token
    history_max_len = 1000  # 模型記憶的最大token長度
    top_p = 0.9
    temperature = 0.35
    repetition_penalty = 1.0

    model = AutoModelForCausalLM.from_pretrained(
        model_name,
        trust_remote_code=True,
        low_cpu_mem_usage=True,
        torch_dtype=torch.float16,
        device_map='auto'
    ).to(device).eval()
    tokenizer = AutoTokenizer.from_pretrained(
        model_name,
        trust_remote_code=True,
        use_fast=False
    )


    history_token_ids = torch.tensor([[]], dtype=torch.long)

    user_input = input('User：')
    while True:
        input_ids = tokenizer(user_input, return_tensors="pt", add_special_tokens=False).input_ids
        eos_token_id = torch.tensor([[tokenizer.eos_token_id]], dtype=torch.long)
        user_input_ids = torch.concat([input_ids, eos_token_id], dim=1)
        history_token_ids = torch.concat((history_token_ids, user_input_ids), dim=1)
        model_input_ids = history_token_ids[:, -history_max_len:].to(device)
        with torch.no_grad():
            outputs = model.generate(
                input_ids=model_input_ids, max_new_tokens=max_new_tokens, do_sample=True, top_p=top_p,
                temperature=temperature, repetition_penalty=repetition_penalty, eos_token_id=tokenizer.eos_token_id
            )
        model_input_ids_len = model_input_ids.size(1)
        response_ids = outputs[:, model_input_ids_len:]
        history_token_ids = torch.concat((history_token_ids, response_ids.cpu()), dim=1)
        response = tokenizer.batch_decode(response_ids)
        print("Bot：" + response[0].strip().replace(tokenizer.eos_token, ""))
        user_input = input('User：')


if __name__ == '__main__':
    main()

📄 許可證

本項目採用OpenRail許可證。

📦 相關信息