Views
No views yet
woohello/olmo3-190m-zh-nano 继续 SFT，
学习指令遵循能力，从"续写文本"转向"扮演 assistant 回答"。woohello/olmo3-190m-zh-nano (26M, OLMo3 arch, SDPA)cmz1024/llm101-olmo3-zh-demo-data/sft/sft_t2t_mini.jsonl (对话格式)1from transformers import AutoModelForCausalLM, AutoTokenizer
2model = AutoModelForCausalLM.from_pretrained("woohello/olmo3-190m-zh-sft", attn_implementation="sdpa")
3tok = AutoTokenizer.from_pretrained("woohello/olmo3-190m-zh-sft")
4
5messages = [{"role": "user", "content": "你好"}]
6input_ids = tok.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
7out = model.generate(input_ids, max_new_tokens=256, do_sample=True, temperature=0.7)
8print(tok.decode(out[0][input_ids.shape[1]:], skip_special_tokens=True))