{ "step": 1, "losses": [ 2.1343601644039154 ], "dataset_offset": 0, "config": { "king_dir": "/root/king_165_hf", "optimizer": "adamw", "lr": 5e-05, "muon_lr": 0.02, "lora_r": 256, "steps": 5000, "grad_accum": 4, "seq_len": 2048, "unfreeze_head": false, "source_dirs": [ "/root/datasets/finewebedu" ], "source_weights": [ 1.0 ] } }