smolm-autoreg-bpe-counterfactual-babylm-pipps_10k-3e-4

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.3363
• Accuracy: 0.4103

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: 0.0003
• 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
3.7369	1.0	18720	3.8674	0.3466
3.4364	2.0	37440	3.5587	0.3765
3.2951	3.0	56160	3.4572	0.3899
3.2109	4.0	74880	3.3845	0.3966
3.1529	5.0	93600	3.3601	0.4002
3.1047	6.0	112320	3.3206	0.4031
3.0672	7.0	131040	3.3101	0.4060
3.0309	8.0	149760	3.3001	0.4068
3.0074	9.0	168480	3.3187	0.4071
2.9719	10.0	187200	3.2961	0.4085
2.9554	11.0	205920	3.2881	0.4098
2.9304	12.0	224640	3.3115	0.4086
2.9115	13.0	243360	3.3158	0.4094
2.8917	14.0	262080	3.2929	0.4107
2.8692	15.0	280800	3.3107	0.4105
2.8477	16.0	299520	3.3204	0.4102
2.8289	17.0	318240	3.3173	0.4099
2.8132	18.0	336960	3.3293	0.4102
2.7963	19.0	355680	3.3299	0.4104
2.7765	20.0	374400	3.3363	0.4103

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.12.0
• Tokenizers 0.14.1