bart-base-summarization-medical_on_cnn-50

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

• Loss: 3.3948
• Rouge1: 0.2474
• Rouge2: 0.0918
• Rougel: 0.1969
• Rougelsum: 0.2198
• Gen Len: 18.166

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: 3e-05
• train_batch_size: 4
• eval_batch_size: 1
• seed: 50
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
2.7028	1.0	1250	3.3697	0.25	0.0901	0.1969	0.2216	18.807
2.5838	2.0	2500	3.3865	0.2496	0.0904	0.196	0.2207	18.586
2.5612	3.0	3750	3.3752	0.2511	0.0928	0.1984	0.2223	18.311
2.5589	4.0	5000	3.3884	0.2508	0.0931	0.1991	0.2222	18.33
2.5524	5.0	6250	3.3936	0.2477	0.0931	0.1967	0.2204	18.159
2.5314	6.0	7500	3.3948	0.2474	0.0918	0.1969	0.2198	18.166

Framework versions

• PEFT 0.12.0
• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1