bert-base-cased-wikitext2

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

• Loss: 6.9846

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
• train_batch_size: 24
• eval_batch_size: 24
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
7.7422	1.0	782	7.1373
7.0302	2.0	1564	6.9972
6.9788	3.0	2346	7.0087

Framework versions

• Transformers 4.11.3
• Pytorch 1.12.0+cu102
• Datasets 1.14.0
• Tokenizers 0.10.3