babylm-balanced_seed-42_1e-3

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.0189
• Accuracy: 0.4204

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 32000
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
6.1706	1.0	1788	4.2554	0.3062
4.0393	2.0	3576	3.7314	0.3478
3.5848	3.0	5364	3.4763	0.3707
3.3537	4.0	7152	3.3407	0.3838
3.264	5.0	8940	3.2649	0.3909
3.1612	6.0	10728	3.2128	0.3958
3.0966	7.0	12516	3.1814	0.3986
3.0531	8.0	14304	3.1567	0.4013
3.0223	9.0	16092	3.1409	0.4034
2.9775	10.0	17880	3.1277	0.4050
2.9579	11.0	19668	3.1193	0.4057
2.9468	12.0	21456	3.1134	0.4066
2.9402	13.0	23244	3.1062	0.4073
2.936	14.0	25032	3.1064	0.4074
2.8948	15.0	26820	3.1046	0.4079
2.8996	16.0	28608	3.0992	0.4085
2.9054	17.0	30396	3.0954	0.4086
2.907	18.0	32184	3.0909	0.4096
2.853	19.0	33972	3.0407	0.4160
2.6995	20.0	35760	3.0189	0.4204

Framework versions

• Transformers 4.45.1
• Pytorch 2.5.1+cu124
• Datasets 2.19.1
• Tokenizers 0.20.0