smolm-autoreg-bpe-counterfactual-babylm-only_other_det_removal-1e-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: 3.4247
• Accuracy: 0.4065

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: 0.0001
• 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
4.0532	1.0	18597	4.2579	0.3085
3.566	2.0	37194	3.7605	0.3620
3.3886	3.0	55791	3.5962	0.3806
3.2899	4.0	74388	3.5175	0.3894
3.2214	5.0	92985	3.4618	0.3939
3.1702	6.0	111582	3.4252	0.3979
3.1294	7.0	130179	3.4255	0.3995
3.0899	8.0	148776	3.4190	0.4010
3.0639	9.0	167373	3.4041	0.4027
3.0329	10.0	185970	3.4231	0.4029
3.0093	11.0	204567	3.4100	0.4045
2.9859	12.0	223164	3.4097	0.4049
2.9662	13.0	241761	3.4043	0.4053
2.9424	14.0	260358	3.4046	0.4057
2.928	15.0	278955	3.4079	0.4059
2.908	16.0	297552	3.4119	0.4061
2.8912	17.0	316149	3.4119	0.4062
2.8716	18.0	334746	3.4159	0.4064
2.8589	19.0	353343	3.4223	0.4065
2.8424	20.0	371940	3.4247	0.4065

Framework versions

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