smolm-autoreg-bpe-babylm-seed_1024-1e-3

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.4044
• Accuracy: 0.4112

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: 128
• seed: 1024
• 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
3.5972	1.0	18595	3.7848	0.3596
3.3783	2.0	37190	3.5924	0.3811
3.2531	3.0	55785	3.4617	0.3919
3.1768	4.0	74380	3.4215	0.3983
3.1209	5.0	92975	3.3891	0.4028
3.0769	6.0	111570	3.3921	0.4039
3.0421	7.0	130165	3.3538	0.4070
3.0089	8.0	148760	3.3607	0.4076
2.9842	9.0	167355	3.3492	0.4094
2.9595	10.0	185950	3.3596	0.4097
2.9382	11.0	204545	3.3544	0.4101
2.9125	12.0	223140	3.3588	0.4107
2.8929	13.0	241735	3.3648	0.4109
2.8689	14.0	260330	3.3662	0.4110
2.8494	15.0	278925	3.3657	0.4116
2.8316	16.0	297520	3.3748	0.4114
2.8151	17.0	316115	3.3865	0.4114
2.797	18.0	334710	3.3825	0.4115
2.7759	19.0	353305	3.4028	0.4111
2.7566	20.0	371900	3.4044	0.4112

Framework versions

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