smolm-autoreg-bpe-counterfactual-babylm-keys_to_pipps_2913-1e-4

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

• Loss: 3.3547
• Accuracy: 0.4065

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: 0.0001
• 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
4.0518	1.0	18629	4.2148	0.3103
3.5731	2.0	37258	3.7002	0.3628
3.3939	3.0	55887	3.5428	0.3792
3.2906	4.0	74516	3.4673	0.3886
3.2245	5.0	93145	3.4056	0.3939
3.1727	6.0	111774	3.4006	0.3969
3.1301	7.0	130403	3.3740	0.3995
3.0954	8.0	149032	3.3669	0.4005
3.0675	9.0	167661	3.3578	0.4022
3.0422	10.0	186290	3.3458	0.4028
3.011	11.0	204919	3.3474	0.4041
2.9957	12.0	223548	3.3452	0.4046
2.9729	13.0	242177	3.3344	0.4055
2.9501	14.0	260806	3.3348	0.4064
2.9303	15.0	279435	3.3385	0.4056
2.91	16.0	298064	3.3449	0.4063
2.8962	17.0	316693	3.3472	0.4060
2.8792	18.0	335322	3.3533	0.4061
2.867	19.0	353951	3.3526	0.4063
2.8451	20.0	372580	3.3547	0.4065

Framework versions

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