Bart_mediasum

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

• Loss: 1.9021
• Rouge1: 0.3236
• Rouge2: 0.1651
• Rougel: 0.2953
• Rougelsum: 0.2953
• Gen Len: 15.7946
• Precision: 0.8858
• Recall: 0.8739
• F1: 0.8795

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: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len	Precision	Recall	F1
2.1171	1.0	4621	2.0135	0.3138	0.1556	0.2853	0.2853	16.4704	0.8836	0.8717	0.8773
1.9804	2.0	9242	1.9440	0.3147	0.1581	0.2864	0.2866	16.2207	0.8831	0.8725	0.8775
1.8971	3.0	13863	1.9157	0.3209	0.1638	0.2925	0.2926	15.4676	0.8857	0.8733	0.8792
1.8449	4.0	18484	1.9021	0.3236	0.1651	0.2953	0.2953	15.7946	0.8858	0.8739	0.8795

Framework versions

• Transformers 4.36.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.15.0