bart-base-finetuned-xsum

This model is a fine-tuned version of facebook/bart-base on the xsum dataset. It achieves the following results on the evaluation set:

• Loss: 2.1755
• Rouge1: 34.6293
• Rouge2: 13.4749
• Rougel: 28.2616
• Rougelsum: 28.2553

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: 5.6e-05
• train_batch_size: 10
• eval_batch_size: 10
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
2.4765	1.0	1000	2.0873	33.9596	12.722	27.4135	27.4062
1.9854	2.0	2000	2.0802	33.6802	12.8965	27.4061	27.4064
1.6677	3.0	3000	2.0998	34.2038	13.1362	27.8808	27.8806
1.4313	4.0	4000	2.1404	34.8491	13.4154	28.2768	28.2702
1.275	5.0	5000	2.1755	34.6293	13.4749	28.2616	28.2553

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2