smolm-autoreg-bpe-counterfactual_babylm_aann_high_variability_adj-seed_211-1e-3

This model was trained from scratch on the kanishka/counterfactual_babylm_aann_high_variability_adj dataset. It achieves the following results on the evaluation set:

• Loss: 3.4027
• 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: 211
• 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
3.5971	1.0	18594	3.7890	0.3595
3.3883	2.0	37188	3.5798	0.3798
3.2541	3.0	55782	3.4594	0.3934
3.1755	4.0	74376	3.4414	0.3975
3.1257	5.0	92970	3.4015	0.4010
3.0777	6.0	111564	3.3917	0.4033
3.0464	7.0	130158	3.3927	0.4057
3.011	8.0	148752	3.3488	0.4094
2.9895	9.0	167346	3.3439	0.4106
2.9591	10.0	185940	3.3509	0.4106
2.9354	11.0	204534	3.3639	0.4102
2.9143	12.0	223128	3.3484	0.4118
2.8939	13.0	241722	3.3607	0.4116
2.8757	14.0	260316	3.3747	0.4107
2.8497	15.0	278910	3.3673	0.4122
2.8279	16.0	297504	3.3792	0.4118
2.8097	17.0	316098	3.3776	0.4125
2.7961	18.0	334692	3.3961	0.4113
2.7712	19.0	353286	3.4010	0.4115
2.7569	20.0	371880	3.4027	0.4116

Framework versions

• Transformers 4.38.0
• Pytorch 2.3.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.2