Modelpublic

shanhy/xlm-roberta-base_latin_original_amh-hau-eng_train_loss

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes8downloads

Model Card

xlm-roberta-baselatinoriginalamh-hau-engtrain_loss

This model is a fine-tuned version of FacebookAI/xlm-roberta-base on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.0269
—Spearman Corr: 0.7656

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	1.55	200	0.0252	0.7541
0.0498	3.1	400	0.0263	0.7765
0.0243	4.65	600	0.0245	0.7840
0.019	6.2	800	0.0235	0.7867
0.019	7.75	1000	0.0335	0.7707
0.0147	9.3	1200	0.0297	0.7705
0.0118	10.85	1400	0.0267	0.7764
0.0096	12.4	1600	0.0276	0.7745
0.0078	13.95	1800	0.0278	0.7631
0.0078	15.5	2000	0.0288	0.7699
0.0067	17.05	2200	0.0249	0.7616
0.0062	18.6	2400	0.0269	0.7656

Framework versions

—Transformers 4.37.2
—Pytorch 2.2.0+cu121
—Datasets 2.17.0
—Tokenizers 0.15.2