bert_uncased_L-4_H-256_A-4_stsb

This model is a fine-tuned version of google/bert_uncased_L-4_H-256_A-4 on the GLUE STSB dataset. It achieves the following results on the evaluation set:

• Loss: 0.6283
• Pearson: 0.8545
• Spearmanr: 0.8542
• Combined Score: 0.8543

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: 256
• eval_batch_size: 256
• seed: 10
• 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Pearson	Spearmanr	Combined Score
5.5773	1.0	23	2.7412	0.3845	0.3343	0.3594
2.5793	2.0	46	1.9158	0.7727	0.7557	0.7642
1.5767	3.0	69	0.9541	0.7706	0.7473	0.7590
0.9474	4.0	92	0.7628	0.8133	0.8070	0.8101
0.7258	5.0	115	0.6785	0.8383	0.8429	0.8406
0.6162	6.0	138	0.6756	0.8436	0.8439	0.8437
0.5455	7.0	161	0.6391	0.8480	0.8504	0.8492
0.4912	8.0	184	0.6582	0.8461	0.8472	0.8466
0.4443	9.0	207	0.6561	0.8472	0.8482	0.8477
0.3995	10.0	230	0.6429	0.8504	0.8503	0.8503
0.3689	11.0	253	0.6283	0.8545	0.8542	0.8543
0.3418	12.0	276	0.6592	0.8520	0.8520	0.8520
0.3302	13.0	299	0.6507	0.8524	0.8530	0.8527
0.319	14.0	322	0.6484	0.8528	0.8526	0.8527
0.2863	15.0	345	0.6397	0.8526	0.8527	0.8526
0.2774	16.0	368	0.6379	0.8559	0.8555	0.8557

Framework versions

• Transformers 4.46.3
• Pytorch 2.2.1+cu118
• Datasets 2.17.0
• Tokenizers 0.20.3