learning_rate: 0.0018
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.95) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 0.05
num_epochs: 1.0

Training Loss	Epoch	Step	Validation Loss
No log	0	0	6.9295
2.7296	0.0500	2014	2.7377
2.5903	0.1000	4028	2.6005
2.5483	0.1500	6042	2.5630
2.5271	0.2000	8056	2.5443
2.5159	0.2500	10070	2.5308
2.5074	0.3000	12084	2.5240
2.4873	0.3500	14098	2.5074
2.4783	0.4000	16112	2.5014
2.4750	0.4500	18126	2.4933
2.4591	0.5000	20140	2.4796
2.4544	0.5500	22154	2.4691
2.4212	0.6000	24168	2.4539
2.4337	0.6500	26182	2.4414
2.4154	0.7000	28196	2.4317
2.3901	0.7500	30210	2.4167
2.3859	0.8000	32224	2.4075
2.3800	0.8500	34238	2.3973
2.3712	0.9000	36252	2.3907
2.3654	0.9500	38266	2.3869
2.3603	1.0	40278	2.3860

stage-babylm/llama-64-1L

llama-64-1L