smolm-autoreg-bpe-counterfactual-babylm-only_other_det_removal-seed_1024-1e-3

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: 3.4160
• Accuracy: 0.4105

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: 1024
• 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
3.6007	1.0	18597	3.8014	0.3575
3.3846	2.0	37194	3.5881	0.3796
3.2609	3.0	55791	3.4855	0.3918
3.1804	4.0	74388	3.4168	0.3979
3.1278	5.0	92985	3.4013	0.4018
3.081	6.0	111582	3.3683	0.4041
3.0471	7.0	130179	3.3773	0.4055
3.0189	8.0	148776	3.3797	0.4069
2.988	9.0	167373	3.3716	0.4074
2.9624	10.0	185970	3.3675	0.4088
2.9372	11.0	204567	3.3803	0.4093
2.9153	12.0	223164	3.3654	0.4096
2.8939	13.0	241761	3.3777	0.4098
2.8704	14.0	260358	3.3811	0.4103
2.8503	15.0	278955	3.3847	0.4102
2.8343	16.0	297552	3.3952	0.4100
2.8131	17.0	316149	3.4062	0.4103
2.7975	18.0	334746	3.4120	0.4102
2.7753	19.0	353343	3.4110	0.4105
2.7567	20.0	371940	3.4160	0.4105

Framework versions

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