opt-babylm2-clean-spacy-32k_seed-42_1e-3

This model was trained from scratch on the kanishka/babylm2-clean-spacy dataset. It achieves the following results on the evaluation set:

• Loss: 3.0380
• Accuracy: 0.4233

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: 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
3.26	1.0	15543	3.3136	0.3831
3.0629	2.0	31086	3.1530	0.3996
2.9355	3.0	46629	3.0425	0.4114
2.8499	4.0	62172	2.9935	0.4172
2.7951	5.0	77715	2.9681	0.4207
2.7489	6.0	93258	2.9527	0.4226
2.7031	7.0	108801	2.9438	0.4242
2.6642	8.0	124344	2.9425	0.4251
2.6242	9.0	139887	2.9441	0.4255
2.5991	10.0	155430	2.9460	0.4258
2.5731	11.0	170973	2.9496	0.4259
2.5387	12.0	186516	2.9571	0.4259
2.5106	13.0	202059	2.9727	0.4256
2.4846	14.0	217602	2.9750	0.4257
2.4638	15.0	233145	2.9852	0.4250
2.4351	16.0	248688	2.9968	0.4247
2.4119	17.0	264231	3.0044	0.4244
2.3894	18.0	279774	3.0132	0.4241
2.3569	19.0	295317	3.0270	0.4237
2.3375	20.0	310860	3.0380	0.4233

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0