smolm-autoreg-bpe-counterfactual-babylm-keys_to_pipps_all-1e-3

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

• Loss: 3.3449
• Accuracy: 0.4108

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: 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
3.6001	1.0	19014	3.7251	0.3609
3.3959	2.0	38028	3.5040	0.3820
3.2662	3.0	57042	3.4255	0.3932
3.1892	4.0	76056	3.3537	0.3992
3.134	5.0	95070	3.3279	0.4033
3.0941	6.0	114084	3.3172	0.4041
3.0584	7.0	133098	3.3045	0.4065
3.026	8.0	152112	3.3046	0.4071
2.9961	9.0	171126	3.2926	0.4084
2.974	10.0	190140	3.2889	0.4090
2.9516	11.0	209154	3.3076	0.4090
2.9292	12.0	228168	3.2888	0.4104
2.9071	13.0	247182	3.2999	0.4101
2.8845	14.0	266196	3.3064	0.4107
2.8669	15.0	285210	3.3155	0.4104
2.8505	16.0	304224	3.3300	0.4105
2.8318	17.0	323238	3.3195	0.4110
2.8155	18.0	342252	3.3315	0.4109
2.7914	19.0	361266	3.3386	0.4109
2.7769	20.0	380280	3.3449	0.4108

Framework versions

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