kanishka/smolm-autoreg-bpe-counterfactual_babylm_aann_low_variability_numeral-seed_1024-1e-3

smolm-autoreg-bpe-counterfactual_babylm_aann_low_variability_numeral_1024-1e-3

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

• Loss: 3.3975
• Accuracy: 0.4113

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
3.598	1.0	18593	3.8055	0.3586
3.3819	2.0	37186	3.5620	0.3807
3.2622	3.0	55779	3.4625	0.3929
3.1788	4.0	74372	3.4131	0.3984
3.1251	5.0	92965	3.3859	0.4022
3.0765	6.0	111558	3.3707	0.4050
3.0458	7.0	130151	3.3522	0.4077
3.0116	8.0	148744	3.3560	0.4080
2.9832	9.0	167337	3.3773	0.4080
2.9565	10.0	185930	3.3513	0.4098
2.9381	11.0	204523	3.3482	0.4096
2.9171	12.0	223116	3.3354	0.4118
2.893	13.0	241709	3.3519	0.4111
2.876	14.0	260302	3.3632	0.4109
2.8472	15.0	278895	3.3530	0.4120
2.8335	16.0	297488	3.3727	0.4113
2.812	17.0	316081	3.3814	0.4110
2.7944	18.0	334674	3.3788	0.4119
2.7761	19.0	353267	3.3925	0.4114
2.7606	20.0	371860	3.3975	0.4113

Framework versions

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