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

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

• Loss: 3.4419
• Accuracy: 0.4081

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
3.7368	1.0	18597	3.9162	0.3452
3.4348	2.0	37194	3.6327	0.3748
3.2919	3.0	55791	3.5084	0.3900
3.2086	4.0	74388	3.4502	0.3956
3.1474	5.0	92985	3.4235	0.3995
3.1012	6.0	111582	3.4031	0.4020
3.0638	7.0	130179	3.4128	0.4030
3.0262	8.0	148776	3.3998	0.4046
3.0016	9.0	167373	3.3731	0.4070
2.9715	10.0	185970	3.4058	0.4062
2.9481	11.0	204567	3.3875	0.4069
2.9243	12.0	223164	3.4070	0.4070
2.9047	13.0	241761	3.4015	0.4079
2.8797	14.0	260358	3.4114	0.4077
2.8651	15.0	278955	3.4072	0.4083
2.8434	16.0	297552	3.4240	0.4075
2.8255	17.0	316149	3.4179	0.4083
2.8036	18.0	334746	3.4256	0.4082
2.7888	19.0	353343	3.4363	0.4083
2.7701	20.0	371940	3.4419	0.4081

Framework versions

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