bert_uncased_L-2_H-128_A-2_stsb

This model is a fine-tuned version of google/bert_uncased_L-2_H-128_A-2 on the GLUE STSB dataset. It achieves the following results on the evaluation set:

• Loss: 0.9185
• Pearson: 0.7914
• Spearmanr: 0.8078
• Combined Score: 0.7996

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
8.7461	1.0	23	5.9052	0.2839	0.2249	0.2544
6.3503	2.0	46	4.1500	0.4510	0.3995	0.4253
4.6275	3.0	69	3.1621	0.5516	0.5407	0.5462
3.6391	4.0	92	2.6588	0.6168	0.6455	0.6311
3.0189	5.0	115	2.3421	0.6722	0.7192	0.6957
2.59	6.0	138	2.0467	0.6785	0.7346	0.7066
2.172	7.0	161	1.6885	0.6686	0.6274	0.6480
1.7948	8.0	184	1.4313	0.6900	0.6565	0.6733
1.5153	9.0	207	1.2854	0.7049	0.7020	0.7035
1.3213	10.0	230	1.1953	0.7136	0.7304	0.7220
1.1482	11.0	253	1.1937	0.7066	0.7005	0.7035
1.0318	12.0	276	1.0680	0.7379	0.7727	0.7553
0.9444	13.0	299	1.0875	0.7445	0.7877	0.7661
0.8957	14.0	322	1.0566	0.7515	0.7869	0.7692
0.8101	15.0	345	1.0417	0.7613	0.7947	0.7780
0.7743	16.0	368	0.9960	0.7708	0.7945	0.7827
0.7407	17.0	391	0.9344	0.7847	0.8062	0.7954
0.6842	18.0	414	0.9185	0.7914	0.8078	0.7996
0.6628	19.0	437	0.9989	0.7836	0.7979	0.7907
0.6402	20.0	460	0.9199	0.7952	0.8082	0.8017
0.6215	21.0	483	0.9276	0.7954	0.8100	0.8027
0.6069	22.0	506	0.9503	0.7956	0.8078	0.8017
0.6101	23.0	529	0.9789	0.7972	0.8134	0.8053

Framework versions

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