来源

https://huggingface.co/violeteverisland/Neko-Chat

🐾 Neko-Chat · 猫娘中文模型

基于 DeepSeek-R1-Distill-Qwen-1.5B 微调的中文轻量对话模型，自带猫娘口癖与亲昵风格。

📌 作者

violeteverisland
GitHub: github.com/violeteverisland
B 站: space.bilibili.com/631367902

🧪 训练信息

• Base: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
• 方法: LoRA → merge & unload
• 数据: 自建 500 条中文短句风格语料（MIT 释出）增强到1w条数据
• 模板: qwen / deepseekr1
• 场景: 角色扮演、闲聊、轻量陪伴

🚀 快速开始

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("violeteverisland/Neko-Chat", device_map="auto", trust_remote_code=True)
tok   = AutoTokenizer.from_pretrained("violeteverisland/Neko-Chat", trust_remote_code=True)

prompt = "主人今天想做什么？"
inputs = tok(prompt, return_tensors="pt").to(model.device)
out    = model.generate(**inputs, max_new_tokens=60, do_sample=True, top_p=0.8)
print(tok.decode(out[0], skip_special_tokens=True))

提示

如果git了本仓库，可以直接采用下面的代码，用来跟猫娘对话

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_path = r"your-model-path"  # ← 改成你的
device = "cuda" if torch.cuda.is_available() else "cpu"

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 chat(prompt, history=None):
    if history is None:
        history = []
 
    text = tokenizer.apply_chat_template(history + [{"role": "user", "content": prompt}], tokenize=False)
    inputs = tokenizer(text, return_tensors="pt").to(device)
    outputs = model.generate(
        **inputs,
        max_new_tokens=512,
        do_sample=True,
        temperature=0.7,
        top_p=0.9,
        repetition_penalty=1.05
    )
    answer = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
    history.append({"role": "user", "content": prompt})
    history.append({"role": "assistant", "content": answer})
    return answer, history

if __name__ == "__main__":
    hist = []
    while True:
        q = input(">>> ")
        if q.lower() in {"exit", "quit"}:
            break
        resp, hist = chat(q, hist)
        print("Bot:", resp)