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

smolm-autoreg-bpe-counterfactual_babylm_aann_low_variability_adj_1024-1e-3

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

• Loss: 3.4324
• Accuracy: 0.4098

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
3.5981	1.0	18593	3.7830	0.3598
3.3835	2.0	37186	3.6126	0.3791
3.255	3.0	55779	3.4609	0.3922
3.1823	4.0	74372	3.4186	0.3985
3.1272	5.0	92965	3.3987	0.4029
3.0802	6.0	111558	3.3908	0.4045
3.0447	7.0	130151	3.3816	0.4063
3.0171	8.0	148744	3.3978	0.4066
2.9869	9.0	167337	3.3734	0.4079
2.9626	10.0	185930	3.3720	0.4088
2.9359	11.0	204523	3.3878	0.4095
2.9153	12.0	223116	3.3868	0.4092
2.8938	13.0	241709	3.4118	0.4098
2.8738	14.0	260302	3.3966	0.4097
2.8542	15.0	278895	3.3984	0.4097
2.8341	16.0	297488	3.4036	0.4097
2.8166	17.0	316081	3.4160	0.4100
2.7931	18.0	334674	3.4189	0.4100
2.7745	19.0	353267	3.4242	0.4099
2.7586	20.0	371860	3.4324	0.4098

Framework versions

• Transformers 4.41.0
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.19.1