smolm-autoreg-bpe-counterfactual-babylm-pipps_and_keys_to_it_all_10k-3e-4

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

• Loss: 3.3416
• Accuracy: 0.4114

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: 0.0003
• 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
3.7439	1.0	18844	3.8602	0.3475
3.4436	2.0	37688	3.5370	0.3777
3.2979	3.0	56532	3.3990	0.3927
3.2129	4.0	75376	3.3575	0.3992
3.1532	5.0	94220	3.3300	0.4014
3.1098	6.0	113064	3.3082	0.4056
3.0691	7.0	131908	3.2938	0.4069
3.042	8.0	150752	3.2975	0.4077
3.0098	9.0	169596	3.2770	0.4112
2.9839	10.0	188440	3.2937	0.4114
2.9607	11.0	207284	3.2879	0.4114
2.94	12.0	226128	3.2938	0.4115
2.9154	13.0	244972	3.3142	0.4101
2.8939	14.0	263816	3.2931	0.4124
2.8771	15.0	282660	3.3156	0.4114
2.8566	16.0	301504	3.3241	0.4112
2.8321	17.0	320348	3.3228	0.4120
2.8173	18.0	339192	3.3250	0.4116
2.7989	19.0	358036	3.3380	0.4114
2.7807	20.0	376880	3.3416	0.4114

Framework versions

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