fine-tuned-16384-pubmed

This model is a fine-tuned version of ccdv/lsg-bart-base-16384-pubmed on the pubmed-summarization dataset. It achieves the following results on the evaluation set:

• Loss: 0.3719
• Rouge1: 0.4602
• Rouge2: 0.2253
• Rougel: 0.2911
• Rougelsum: 0.4283

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
12.7776	0.2667	50	9.5450	0.4552	0.2219	0.2848	0.4234
7.2062	0.5333	100	4.6159	0.4564	0.2215	0.2874	0.4241
4.0511	0.8	150	2.7224	0.4615	0.2222	0.2914	0.4287
2.657	1.0667	200	1.7347	0.4597	0.2226	0.2921	0.4273
1.7104	1.3333	250	1.0021	0.4583	0.2231	0.2918	0.4261
0.9336	1.6	300	0.5423	0.4586	0.2228	0.2905	0.4259
0.4902	1.8667	350	0.3719	0.4602	0.2253	0.2911	0.4283

Framework versions

• Transformers 4.43.3
• Pytorch 2.0.0
• Datasets 2.15.0
• Tokenizers 0.19.1