kanishka/smolm-autoreg-bpe-counterfactual_babylm_aann_dtanns-seed_211-1e-4

smolm-autoreg-bpe-counterfactual_babylm_aanns_dtanns-seed_211-1e-4

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

• Loss: 3.4323
• Accuracy: 0.4055

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: 211
• 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
4.0524	1.0	18595	4.2811	0.3084
3.5716	2.0	37190	3.7750	0.3615
3.3913	3.0	55785	3.6008	0.3779
3.2921	4.0	74380	3.5102	0.3883
3.2283	5.0	92975	3.4789	0.3930
3.1673	6.0	111570	3.4379	0.3962
3.127	7.0	130165	3.4175	0.3988
3.0935	8.0	148760	3.4272	0.3998
3.0679	9.0	167355	3.4150	0.4011
3.0402	10.0	185950	3.4148	0.4023
3.0082	11.0	204545	3.4169	0.4029
2.988	12.0	223140	3.4130	0.4035
2.967	13.0	241735	3.3969	0.4045
2.9457	14.0	260330	3.4080	0.4049
2.9268	15.0	278925	3.4129	0.4045
2.911	16.0	297520	3.4159	0.4047
2.8964	17.0	316115	3.4221	0.4051
2.8758	18.0	334710	3.4286	0.4054
2.858	19.0	353305	3.4265	0.4054
2.8493	20.0	371900	3.4323	0.4055

Framework versions

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