smolm-autoreg-bpe-counterfactual_babylm_aann_dtanns-1e-3

This model was trained from scratch on the kanishka/counterfactual_babylm_aann_dtanns dataset. It achieves the following results on the evaluation set:

• Loss: 3.4127
• Accuracy: 0.4103

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
• 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
3.6039	1.0	18595	3.8025	0.3584
3.3897	2.0	37190	3.5640	0.3815
3.2612	3.0	55785	3.4620	0.3929
3.1771	4.0	74380	3.4317	0.3984
3.1255	5.0	92975	3.3973	0.4021
3.0763	6.0	111570	3.4013	0.4047
3.0449	7.0	130165	3.3915	0.4064
3.015	8.0	148760	3.3709	0.4065
2.9863	9.0	167355	3.3587	0.4082
2.9606	10.0	185950	3.3685	0.4091
2.9353	11.0	204545	3.3655	0.4091
2.9156	12.0	223140	3.3699	0.4095
2.8967	13.0	241735	3.3612	0.4106
2.87	14.0	260330	3.3666	0.4104
2.8519	15.0	278925	3.3639	0.4113
2.8296	16.0	297520	3.3780	0.4110
2.8137	17.0	316115	3.3897	0.4106
2.7928	18.0	334710	3.4014	0.4103
2.7756	19.0	353305	3.4025	0.4107
2.7602	20.0	371900	3.4127	0.4103

Framework versions

• Transformers 4.38.0
• Pytorch 2.3.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.2