smolm-autoreg-bpe-counterfactual-babylm-pipps-random_removal-1e-3

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

• Loss: 3.3829
• Accuracy: 0.4120

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
3.6058	1.0	18592	3.8079	0.3582
3.3918	2.0	37184	3.5864	0.3803
3.264	3.0	55776	3.4837	0.3930
3.1794	4.0	74368	3.4301	0.3984
3.1239	5.0	92960	3.3843	0.4023
3.0814	6.0	111552	3.3626	0.4045
3.0416	7.0	130144	3.3471	0.4076
3.0128	8.0	148736	3.3522	0.4079
2.9879	9.0	167328	3.3497	0.4087
2.9616	10.0	185920	3.3193	0.4123
2.941	11.0	204512	3.3381	0.4113
2.9156	12.0	223104	3.3479	0.4114
2.8946	13.0	241696	3.3280	0.4130
2.8744	14.0	260288	3.3445	0.4123
2.8532	15.0	278880	3.3571	0.4119
2.831	16.0	297472	3.3629	0.4122
2.8168	17.0	316064	3.3629	0.4121
2.7943	18.0	334656	3.3743	0.4119
2.7777	19.0	353248	3.3781	0.4121
2.7631	20.0	371840	3.3829	0.4120

Framework versions

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