Modelpublic

Shijia/furina_seed42_eng_kin_amh_cross_latin_2e-05

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes9downloads

Model Card

furinaseed42engkinamhcrosslatin_2e-05

This model is a fine-tuned version of yihongLiu/furina on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.0204
—Spearman Corr: 0.7440

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: 2e-05
—trainbatchsize: 32
—evalbatchsize: 128
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 64
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 30
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Spearman Corr
No log	0.59	200	0.0262	0.6019
No log	1.17	400	0.0231	0.6882
No log	1.76	600	0.0287	0.6986
0.0478	2.35	800	0.0239	0.7144
0.0478	2.93	1000	0.0201	0.7331
0.0478	3.52	1200	0.0205	0.7360
0.0206	4.11	1400	0.0203	0.7431
0.0206	4.69	1600	0.0209	0.7377
0.0206	5.28	1800	0.0197	0.7457
0.0206	5.87	2000	0.0203	0.7442
0.0141	6.45	2200	0.0198	0.7482
0.0141	7.04	2400	0.0208	0.7452
0.0141	7.62	2600	0.0206	0.7411
0.0104	8.21	2800	0.0215	0.7371
0.0104	8.8	3000	0.0198	0.7441
0.0104	9.38	3200	0.0210	0.7422
0.0104	9.97	3400	0.0204	0.7440

Framework versions

—Transformers 4.37.2
—Pytorch 2.1.0+cu121
—Datasets 2.17.0
—Tokenizers 0.15.2