Model Card for Doge-20M-Medical-SFT

This model is a fine-tuned version of wubingheng/Doge-20M-Chinese. It has been trained using TRL.

Quick start

from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig, TextStreamer
tokenizer = AutoTokenizer.from_pretrained("wubingheng/Doge-20M-Medical-SFT")
model = AutoModelForCausalLM.from_pretrained("wubingheng/Doge-20M-Medical-SFT", trust_remote_code=True)
generation_config = GenerationConfig(
    max_new_tokens=40,
    min_new_tokens=1,
    num_beams=1,
    eos_token_id=[tokenizer.eos_token_id],
    stop_strings=[tokenizer.eos_token],
    early_stopping=False,
    use_cache=True,
    do_sample=True,
    temperature=0.95,
    repetition_penalty=1.0,
)
steamer = TextStreamer(tokenizer=tokenizer, skip_prompt=True)

system_prompt = """
    你是一个医学助手，能够回答用户提出的医学问题。请根据用户的问题，给出准确的医学建议和解答。

""".strip()

prompt = "肝癌术后饮食注意事项有哪些" 
conversation = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": prompt},
]
inputs = tokenizer.apply_chat_template(
    conversation=conversation,
    tokenize=True,
    return_tensors="pt",
)
print(prompt)

output = model.generate(
    inputs, 
    tokenizer=tokenizer,
    generation_config=generation_config, 
    streamer=steamer
)

Training procedure

This model was trained with SFT.

Framework versions

• TRL: 0.15.2
• Transformers: 4.51.2
• Pytorch: 2.6.0+cu126
• Datasets: 3.2.0
• Tokenizers: 0.21.0

Citations

@misc{smalldoges,
  title={SmallDoges: A Family of Dynamic UltraFast Small Language Models}, 
  author={Jingze, Shi and Yifan, Wu and Bingheng, Wu and Yuyu, Luo},
  year={2025},
  month={March},
  url={https://github.com/SmallDoges/small-doge}
}