Modelpublic

Shijia/furina_seed42_eng_amh_hau_roman

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes10downloads

Model Card

furinaseed42engamhhau_roman

This model is a fine-tuned version of yihongLiu/furina on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.0233
—Spearman Corr: 0.7621

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	0.58	200	0.0306	0.6454
No log	1.15	400	0.0353	0.6854
No log	1.73	600	0.0298	0.7055
0.0458	2.3	800	0.0307	0.7105
0.0458	2.88	1000	0.0263	0.7299
0.0458	3.45	1200	0.0273	0.7357
0.0222	4.03	1400	0.0255	0.7374
0.0222	4.6	1600	0.0268	0.7398
0.0222	5.18	1800	0.0316	0.7371
0.0222	5.76	2000	0.0245	0.7445
0.0155	6.33	2200	0.0264	0.7484
0.0155	6.91	2400	0.0311	0.7549
0.0155	7.48	2600	0.0223	0.7585
0.0112	8.06	2800	0.0257	0.7483
0.0112	8.63	3000	0.0240	0.7507
0.0112	9.21	3200	0.0275	0.7609
0.0112	9.78	3400	0.0265	0.7565
0.0086	10.36	3600	0.0250	0.7534
0.0086	10.94	3800	0.0285	0.7577
0.0086	11.51	4000	0.0225	0.7625
0.007	12.09	4200	0.0233	0.7621

Framework versions

—Transformers 4.37.2
—Pytorch 2.1.0+cu121
—Datasets 2.17.0
—Tokenizers 0.15.1