smolm-autoreg-bpe-counterfactual-babylm-keys_to_pipps_all-1e-4

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.3725
• Accuracy: 0.4068

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
4.0248	1.0	19014	4.2224	0.3118
3.5712	2.0	38028	3.6678	0.3643
3.3951	3.0	57042	3.5394	0.3804
3.2986	4.0	76056	3.4576	0.3890
3.2317	5.0	95070	3.4292	0.3932
3.1832	6.0	114084	3.3886	0.3968
3.1425	7.0	133098	3.3770	0.3993
3.1062	8.0	152112	3.3730	0.4013
3.0733	9.0	171126	3.3467	0.4030
3.0496	10.0	190140	3.3508	0.4038
3.0261	11.0	209154	3.3645	0.4039
3.003	12.0	228168	3.3508	0.4050
2.9808	13.0	247182	3.3479	0.4059
2.9585	14.0	266196	3.3544	0.4059
2.9413	15.0	285210	3.3535	0.4063
2.9256	16.0	304224	3.3666	0.4058
2.9089	17.0	323238	3.3630	0.4067
2.8949	18.0	342252	3.3693	0.4064
2.8723	19.0	361266	3.3718	0.4067
2.8605	20.0	380280	3.3725	0.4068

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.12.0
• Tokenizers 0.14.1