bbq-distil_bumble_bert

This model is a fine-tuned version of Aubins/distil-bumble-bert on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1032
• Accuracy: 0.9627
• Precision: 0.9432
• Recall: 0.9470
• F1: 0.9451
• Roc Auc: 0.9965

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Roc Auc
0.3899	0.1709	500	0.3561	0.8325	0.7513	0.7559	0.7536	0.9373
0.3563	0.3419	1000	0.3456	0.8429	0.7693	0.7662	0.7678	0.9444
0.3987	0.5128	1500	0.3510	0.8402	0.7658	0.7612	0.7635	0.9424
0.4003	0.6838	2000	0.3447	0.8595	0.7909	0.7957	0.7933	0.9523
0.2942	0.8547	2500	0.3214	0.8660	0.7998	0.8063	0.8031	0.9577
0.288	1.0256	3000	0.3118	0.8774	0.8158	0.8245	0.8201	0.9642
0.2941	1.1966	3500	0.2656	0.8886	0.8303	0.8439	0.8370	0.9715
0.2818	1.3675	4000	0.2618	0.9015	0.8458	0.8676	0.8566	0.9763
0.265	1.5385	4500	0.2281	0.9093	0.8589	0.8764	0.8676	0.9804
0.1927	1.7094	5000	0.1938	0.9297	0.8929	0.9004	0.8966	0.9869
0.1919	1.8803	5500	0.1726	0.9394	0.9038	0.9190	0.9113	0.9902
0.1421	2.0513	6000	0.1578	0.9426	0.9111	0.9206	0.9158	0.9918
0.1481	2.2222	6500	0.1429	0.9464	0.9189	0.9233	0.9211	0.9930
0.1363	2.3932	7000	0.1219	0.9562	0.9317	0.9397	0.9357	0.9948
0.2112	2.5641	7500	0.1173	0.9594	0.9391	0.9412	0.9402	0.9956
0.1424	2.7350	8000	0.1102	0.9604	0.9391	0.9445	0.9418	0.9961
0.1744	2.9060	8500	0.1032	0.9627	0.9432	0.9470	0.9451	0.9965

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0