Modelpublic

Hartunka/bert_base_km_100_v1_stsb

sourceHugging Faceupdated 1y agoView on Hugging Face

0likes12downloads

Model Card

bertbasekm100v1_stsb

This model is a fine-tuned version of Hartunka/bert_base_km_100_v1 on the GLUE STSB dataset. It achieves the following results on the evaluation set:

—Loss: 2.2490
—Pearson: 0.2995
—Spearmanr: 0.2983
—Combined Score: 0.2989

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
—trainbatchsize: 256
—evalbatchsize: 256
—seed: 10
—optimizer: Use adamwtorch with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Pearson	Spearmanr	Combined Score
2.7396	1.0	23	2.5547	0.1514	0.1351	0.1432
1.9555	2.0	46	2.3166	0.1712	0.1497	0.1605
1.749	3.0	69	2.3146	0.2127	0.2000	0.2064
1.3865	4.0	92	2.2490	0.2995	0.2983	0.2989
0.9821	5.0	115	2.7978	0.2457	0.2364	0.2410
0.6935	6.0	138	2.8239	0.2598	0.2516	0.2557
0.4947	7.0	161	2.9618	0.2405	0.2309	0.2357
0.3874	8.0	184	2.7149	0.2566	0.2501	0.2533
0.31	9.0	207	2.5269	0.2768	0.2706	0.2737

Framework versions

—Transformers 4.50.2
—Pytorch 2.2.1+cu121
—Datasets 2.18.0
—Tokenizers 0.21.1