bart-base-summarization-medical_on_cnn-48

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: 3.3893
• Rouge1: 0.2525
• Rouge2: 0.0944
• Rougel: 0.2
• Rougelsum: 0.2242
• Gen Len: 18.451

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: 48
• 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
2.6901	1.0	1250	3.3869	0.2516	0.0884	0.1964	0.2218	19.066
2.6035	2.0	2500	3.3751	0.2516	0.0926	0.1975	0.2231	18.716
2.564	3.0	3750	3.3818	0.2503	0.0926	0.1974	0.2221	18.501
2.5265	4.0	5000	3.3882	0.2505	0.0927	0.1979	0.2219	18.482
2.5207	5.0	6250	3.3881	0.2532	0.0946	0.2005	0.2247	18.394
2.5356	6.0	7500	3.3893	0.2525	0.0944	0.2	0.2242	18.451

Framework versions

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