lm-checkpoints-2026

Storage for fine-tuned language-model checkpoints. Each folder is a full set of merged weights in safetensors format together with the tokenizer and chat template of its base model.

Folder	Base model	License
m01	Qwen/Qwen3.5-9B	Apache-2.0
m02	Qwen/Qwen3.5-9B	Apache-2.0
m03	Qwen/Qwen3.5-9B	Apache-2.0
m04	Qwen/Qwen3.5-9B	Apache-2.0
m05	Qwen/Qwen3.5-9B	Apache-2.0
m06	Qwen/Qwen3.5-9B	Apache-2.0
m07	allenai/Olmo-Hybrid-Instruct-DPO-7B	Apache-2.0
m08	allenai/Olmo-Hybrid-Instruct-DPO-7B	Apache-2.0
m09	Qwen/Qwen3.5-9B	Apache-2.0

The weights in every folder were modified from the listed base model by fine-tuning. The base models are released under the Apache License 2.0, and these checkpoints are distributed under the same license; a copy is in LICENSE. Load a folder with transformers, for example AutoModelForCausalLM.from_pretrained("hyoungjoon/lm-checkpoints-2026", subfolder="m01").