smolm-autoreg-bpe-counterfactual-babylm-random_removal-3e-4

This model was trained from scratch on the kanishka/counterfactual-babylm-random_removal dataset. It achieves the following results on the evaluation set:

• Loss: 3.4045
• Accuracy: 0.4106

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
3.7431	1.0	18586	3.8860	0.3481
3.4382	2.0	37172	3.6288	0.3759
3.2958	3.0	55758	3.5019	0.3891
3.2055	4.0	74344	3.4541	0.3957
3.1533	5.0	92930	3.4028	0.4006
3.1056	6.0	111516	3.3805	0.4035
3.0671	7.0	130102	3.3833	0.4042
3.0331	8.0	148688	3.3669	0.4069
3.0072	9.0	167274	3.3616	0.4083
2.9771	10.0	185860	3.3777	0.4078
2.9534	11.0	204446	3.3741	0.4089
2.9279	12.0	223032	3.3845	0.4092
2.9063	13.0	241618	3.3689	0.4105
2.8913	14.0	260204	3.3711	0.4105
2.8704	15.0	278790	3.3779	0.4099
2.8491	16.0	297376	3.3760	0.4112
2.83	17.0	315962	3.3752	0.4116
2.8136	18.0	334548	3.3912	0.4108
2.7924	19.0	353134	3.3984	0.4107
2.7792	20.0	371720	3.4045	0.4106

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1