learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 195
training_steps: 19500

Training Loss	Epoch	Step	Validation Loss	Accuracy
6.8743	0.1122	200	6.5233	0.1022
6.1033	0.2245	400	6.2006	0.1237
5.9405	0.3367	600	6.0823	0.1329
5.8236	0.4489	800	6.0260	0.1374
5.777	0.5612	1000	5.9901	0.1381
5.7245	0.6734	1200	5.9425	0.1430
5.7086	0.7856	1400	5.9186	0.1472
5.6794	0.8979	1600	5.8993	0.1473
5.6582	1.0101	1800	5.8833	0.1476
5.5798	1.1223	2000	5.8732	0.1492
5.4905	2.2447	4000	5.7809	0.1542
5.3537	3.3670	6000	5.7190	0.1586
5.3199	4.4893	8000	5.6712	0.1601
5.2578	5.6117	10000	5.6395	0.1610
5.2367	6.7340	12000	5.6065	0.1616
5.2088	7.8563	14000	5.5854	0.1632
5.1523	8.9787	16000	5.5658	0.1644
5.1263	10.1010	18000	5.5506	0.1649

alexandertam/babylm-base2f5m-roberta

babylm-base2.5m-roberta