RoBERTa-perigon-news

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

• Loss: 0.9548

Model description

The model was pre-trained for a MLM taskusing over 200K financial news articles obtaind from Perigon https://www.goperigon.com/.

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: 8.7e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.19
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss
1.4872	1.0	5480	1.3355
1.3571	2.0	10960	1.2488
1.3078	3.0	16440	1.2144
1.2425	4.0	21920	1.1634
1.2035	5.0	27400	1.1309
1.157	6.0	32880	1.0941
1.1268	7.0	38360	1.0696
1.098	8.0	43840	1.0466
1.0681	9.0	49320	1.0297
1.0356	10.0	54800	1.0168
1.0194	11.0	60280	1.0011
0.9941	12.0	65760	0.9843
0.981	13.0	71240	0.9716
0.9634	14.0	76720	0.9600
0.9511	15.0	82200	0.9546

Framework versions

• Transformers 4.33.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3