opt-babylm2-20-epochs_seed-42_3e-4

This model was trained from scratch on the kanishka/babylm2-sentence-tokenized dataset. It achieves the following results on the evaluation set:

• Loss: 2.4950
• Accuracy: 0.5193

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: 0.0003
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• 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
2.8017	1.0	21397	2.9043	0.4674
2.5697	2.0	42794	2.6914	0.4903
2.4593	3.0	64191	2.5998	0.5009
2.3962	4.0	85588	2.5532	0.5062
2.3371	5.0	106985	2.5247	0.5100
2.3029	6.0	128382	2.5101	0.5121
2.2663	7.0	149779	2.4970	0.5143
2.2435	8.0	171176	2.4892	0.5155
2.2171	9.0	192573	2.4831	0.5163
2.1902	10.0	213970	2.4811	0.5171
2.1695	11.0	235367	2.4788	0.5177
2.1548	12.0	256764	2.4811	0.5182
2.1307	13.0	278161	2.4788	0.5186
2.1228	14.0	299558	2.4802	0.5188
2.0984	15.0	320955	2.4807	0.5190
2.0845	16.0	342352	2.4828	0.5192
2.0687	17.0	363749	2.4844	0.5193
2.0578	18.0	385146	2.4892	0.5193
2.0413	19.0	406543	2.4918	0.5193
2.0185	20.0	427940	2.4950	0.5193

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0