smolm-autoreg-bpe-counterfactual-babylm-only_other_det_removal-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: 3.4193
• Accuracy: 0.4116

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
3.6043	1.0	18597	3.7893	0.3595
3.3863	2.0	37194	3.5796	0.3811
3.2568	3.0	55791	3.4811	0.3933
3.1802	4.0	74388	3.4316	0.3992
3.1237	5.0	92985	3.3913	0.4033
3.0797	6.0	111582	3.4136	0.4042
3.0447	7.0	130179	3.3948	0.4058
3.0084	8.0	148776	3.3772	0.4079
2.985	9.0	167373	3.3589	0.4101
2.9555	10.0	185970	3.3777	0.4096
2.9324	11.0	204567	3.3606	0.4110
2.9092	12.0	223164	3.3722	0.4112
2.89	13.0	241761	3.3737	0.4114
2.8651	14.0	260358	3.3934	0.4110
2.8499	15.0	278955	3.3911	0.4116
2.8292	16.0	297552	3.3942	0.4114
2.8105	17.0	316149	3.4117	0.4113
2.7877	18.0	334746	3.4073	0.4116
2.773	19.0	353343	3.4169	0.4115
2.7535	20.0	371940	3.4193	0.4116

Framework versions

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