Modelpublic

Shijia/xlm-roberta-base_arb_corr_5e-06

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes7downloads

Model Card

xlm-roberta-basearbcorr_5e-06

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: 0.0254
—Spearman Corr: 0.7597

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-06
—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	0.85	200	0.0234	0.7626
No log	1.69	400	0.0230	0.7567
0.011	2.54	600	0.0229	0.7541
0.011	3.38	800	0.0244	0.7537
0.0114	4.23	1000	0.0252	0.7601
0.0114	5.07	1200	0.0242	0.7650
0.0114	5.92	1400	0.0245	0.7596
0.0098	6.77	1600	0.0255	0.7522
0.0098	7.61	1800	0.0243	0.7595
0.0089	8.46	2000	0.0242	0.7549
0.0089	9.3	2200	0.0247	0.7617
0.0085	10.15	2400	0.0255	0.7578
0.0085	10.99	2600	0.0266	0.7583
0.0085	11.84	2800	0.0254	0.7597

Framework versions

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