opt-babylm2-rewritten-clean-spacy-earlystop_hierarchical_211_age-origin_adj2-bpe_seed-42_1e-3

This model was trained from scratch on the kanishka/babylm2-rewritten-clean-spacy_hierarchical-adj_211_age-origin_adj2-ablation dataset. It achieves the following results on the evaluation set:

• Loss: 2.6855
• Accuracy: 0.4782

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
• gradient_accumulation_steps: 8
• total_train_batch_size: 256
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• 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
4.0984	1.0	2256	3.8023	0.3621
3.443	2.0	4512	3.2925	0.4107
3.1262	3.0	6768	3.0846	0.4309
2.925	4.0	9024	2.9840	0.4408
2.8441	5.0	11280	2.9227	0.4474
2.7863	6.0	13536	2.8835	0.4513
2.7429	7.0	15792	2.8565	0.4544
2.7139	8.0	18048	2.8359	0.4569
2.6913	9.0	20304	2.8223	0.4586
2.6674	10.0	22560	2.8129	0.4597
2.6511	11.0	24816	2.8069	0.4602
2.6505	12.0	27072	2.8003	0.4612
2.6407	13.0	29328	2.7929	0.4620
2.6292	14.0	31584	2.7876	0.4625
2.6134	15.0	33840	2.7690	0.4650
2.5683	16.0	36096	2.7418	0.4681
2.5205	17.0	38352	2.7217	0.4714
2.4641	18.0	40608	2.7012	0.4743
2.3972	19.0	42864	2.6873	0.4769
2.3258	20.0	45120	2.6855	0.4782

Framework versions

• Transformers 4.48.0
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.4