chris0809/tinyLLM-Demo

tinyLLM 0.51B Demo

<!-- tinyllm-showcase --> 中文项目文档 · English project overview · Live demo

<table><tr><td width="50%"><img src="https://github.com/Huanz86251/tinyLLM/raw/main/assets/showcase/tinyllm-thinking-demo.gif" alt="tinyLLM SFT reasoning and tested Python"></td><td width="50%"><img src="https://github.com/Huanz86251/tinyLLM/raw/main/assets/showcase/tinyllm-vlm-demo.gif" alt="tinyLLM VLM image demo"></td></tr><tr><td align="center"><b>SFT reasoning + tested Python</b></td><td align="center"><b>Local VLM image understanding</b></td></tr></table>

This demo is part of a from-scratch 0.51B bilingual small language model project covering SFT, GRPO, on-policy distillation, LoRA, and vision-language alignment. <!-- /tinyllm-showcase -->

This free ZeroGPU Space loads the public tinyLLM 0.51B SFT checkpoint and offers a small streaming chat demo.

The first start is slower while the 0.51B checkpoint is downloaded. The full multimodal demo remains available in the GitHub repository; this small hosted Space focuses on the stable text base. ARC GRPO and IFEval OPD adapters are linked from the app and can be loaded locally with load_lora_pretrained().