Views
No views yet
Qwen/Qwen2.5-3B-Instruct をベースに、1from unsloth import FastLanguageModel
2
3# BaseModek
4base = "Qwen/Qwen2.5-3B-Instruct"
5
6# Load LoRa Adapter
7model, tok = FastLanguageModel.from_pretrained(base, load_in_4bit=True, max_seq_length=1024)
8model.load_adapter("your-username/nanj-qwen2.5-3b-merged")
9
10# example
11msgs = [{"role":"user", "content":"藤浪復活したら阪神どうなる？"}]
12prompt = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)
13
14out = model.generate(
15 **tok(prompt, return_tensors="pt").to(model.device),
16 max_new_tokens=128, do_sample=True, top_p=0.9, temperature=0.8
17)
18print(tok.decode(out[0], skip_special_tokens=True).split(prompt)[-1].strip())
191@misc{vonwerra2022trl,
2 title = {{TRL: Transformer Reinforcement Learning}},
3 author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
4 year = 2020,
5 journal = {GitHub repository},
6 publisher = {GitHub},
7 howpublished = {\url{https://github.com/huggingface/trl}}
8}