mlx_lm.lora
--model mlx-community/Josiefied-Qwen2.5-0.5B-Instruct-abliterated-v1
--train
--data mlx-community/Dolci-Instruct-SFT-No-Tools-100K
--iters 100
--max-seq-length… See the full description on the dataset page:
https://huggingface.co/datasets/mlx-community/Dolci-Instruct-SFT-No-Tools-100K.