from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="sapbot/gemma-4-31b-it-distill-qwen3-0.6b-lora", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Interesting changes in behavior

While Gemma-3n-4b-distill-smollm2-360m-instruct had almost zero changes, because of literally typical behavior, and no knowledge of yourself at all (except system prompt which still guided distilled version that it's HF's model), this one has BIG difference: thinking.

Original Qwen3 0.6B model doesn't have structured reasoning, it just... reasoning. Like humans, no structure. Due to finetuning on Gemma 4 which has structured reasoning, this thinking patterns are now inside of Qwen3 0.6B.

Training procedure

This model was trained with SFT.

Framework versions

PEFT 0.18.1
TRL: 0.23.1
Transformers: 4.57.6
Pytorch: 2.10.0+cu126
Datasets: 4.3.0
Tokenizers: 0.22.2

Citations

Cite TRL as:

bibtex

@misc{vonwerra2022trl,
	title        = {{TRL: Transformer Reinforcement Learning}},
	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},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{https://github.com/huggingface/trl}}
}

Brought to you by sapbot from Romarchive