smolm-autoreg-bpe-counterfactual-babylm-indef-removal-seed_211-1e-4

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

• Loss: 3.4163
• Accuracy: 0.4076

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
4.0497	1.0	18595	4.2852	0.3080
3.565	2.0	37190	3.7283	0.3633
3.3896	3.0	55785	3.5917	0.3804
3.2893	4.0	74380	3.5098	0.3893
3.2194	5.0	92975	3.4496	0.3948
3.1724	6.0	111570	3.4233	0.3982
3.1258	7.0	130165	3.4181	0.4005
3.0958	8.0	148760	3.4253	0.4011
3.0662	9.0	167355	3.4222	0.4023
3.0368	10.0	185950	3.4052	0.4037
3.0083	11.0	204545	3.3978	0.4047
2.9889	12.0	223140	3.3996	0.4056
2.9622	13.0	241735	3.3896	0.4065
2.9419	14.0	260330	3.3847	0.4072
2.9292	15.0	278925	3.4022	0.4065
2.9096	16.0	297520	3.4086	0.4071
2.893	17.0	316115	3.4042	0.4074
2.872	18.0	334710	3.4126	0.4074
2.8563	19.0	353305	3.4111	0.4076
2.8424	20.0	371900	3.4163	0.4076

Framework versions

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