bart-base-finetuned-kaggglenews-fact-corrector-II

This model is a fine-tuned version of facebook/bart-base on the None dataset.

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: 0.0002
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	305	1.5749	27.9313	15.1004	23.3282	25.2336	20.0

Framework versions

• Transformers 4.12.5
• Pytorch 1.10.0+cu102
• Datasets 1.16.1
• Tokenizers 0.10.3