roberta-base-lm-all

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

• Loss: 0.7109

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-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 16

Training results

Training Loss	Epoch	Step	Validation Loss
1.2966	1.0	1194	1.0711
1.0858	2.0	2388	0.9740
1.0055	3.0	3582	0.9273
0.9301	4.0	4776	0.8784
0.9021	5.0	5970	0.8731
0.8479	6.0	7164	0.8406
0.8142	7.0	8358	0.8172
0.7858	8.0	9552	0.8158
0.7529	9.0	10746	0.7922
0.7189	10.0	11940	0.7855
0.7032	11.0	13134	0.7761
0.6795	12.0	14328	0.7549
0.6673	13.0	15522	0.7277
0.6412	14.0	16716	0.7121
0.6321	15.0	17910	0.7168
0.6198	16.0	19104	0.7109

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1