mBART-TextSimp-LT-BatchSize8-lr1e-4

This model is a fine-tuned version of facebook/mbart-large-50 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0744
• Rouge1: 0.7931
• Rouge2: 0.657
• Rougel: 0.7873
• Sacrebleu: 56.7266
• Gen Len: 34.9093

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Sacrebleu	Gen Len
5.5149	1.0	104	3.9238	0.6383	0.4681	0.6268	40.2914	36.6301
0.0989	2.0	209	0.0807	0.6925	0.5318	0.6852	46.8337	34.9093
0.0666	3.0	313	0.0679	0.7014	0.5448	0.6936	47.4249	34.9093
0.0464	4.0	418	0.0765	0.698	0.5344	0.6859	46.7217	34.9093
0.0344	5.0	522	0.0668	0.7604	0.6085	0.7526	49.5391	34.9117
0.0168	6.0	627	0.0652	0.7825	0.6391	0.7763	53.7888	34.9093
0.0116	7.0	731	0.0674	0.7911	0.6558	0.7856	55.8977	34.9093
0.0036	7.96	832	0.0744	0.7931	0.657	0.7873	56.7266	34.9093

Framework versions

• Transformers 4.33.0
• Pytorch 2.1.2+cu121
• Datasets 2.14.4
• Tokenizers 0.13.3