smolm-autoreg-bpe-counterfactual-babylm-pipps_and_keys_to_it_all_10k-1e-4

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

• Loss: 3.3595
• Accuracy: 0.4079

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: 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
4.056	1.0	18844	4.2267	0.3114
3.5768	2.0	37688	3.6840	0.3636
3.3965	3.0	56532	3.5042	0.3817
3.2965	4.0	75376	3.4624	0.3893
3.229	5.0	94220	3.4083	0.3940
3.1807	6.0	113064	3.3809	0.3974
3.1371	7.0	131908	3.3657	0.3999
3.1078	8.0	150752	3.3597	0.4015
3.0741	9.0	169596	3.3538	0.4030
3.0472	10.0	188440	3.3511	0.4047
3.0234	11.0	207284	3.3451	0.4048
3.0026	12.0	226128	3.3451	0.4062
2.9792	13.0	244972	3.3464	0.4066
2.9576	14.0	263816	3.3389	0.4073
2.9415	15.0	282660	3.3515	0.4072
2.9218	16.0	301504	3.3501	0.4073
2.8992	17.0	320348	3.3496	0.4077
2.8867	18.0	339192	3.3493	0.4078
2.8695	19.0	358036	3.3611	0.4077
2.8538	20.0	376880	3.3595	0.4079

Framework versions

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