bart-paraphrase-finetuned-xsum-v3

This model is a fine-tuned version of eugenesiow/bart-paraphrase on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1881
• Rouge1: 99.9251
• Rouge2: 99.9188
• Rougel: 99.9251
• Rougelsum: 99.9251
• Gen Len: 10.17

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	100	0.2702	99.9251	99.9188	99.9251	99.9251	10.38
No log	2.0	200	0.2773	99.9251	99.9188	99.9251	99.9251	11.45
No log	3.0	300	0.2178	99.8148	99.7051	99.8208	99.8148	11.19
No log	4.0	400	0.3649	99.9251	99.9188	99.9251	99.9251	12.32
0.1561	5.0	500	0.2532	99.8957	99.8875	99.8957	99.8918	10.375
0.1561	6.0	600	0.2050	99.9251	99.9188	99.9251	99.9251	11.15
0.1561	7.0	700	0.2364	99.8957	99.8875	99.8957	99.8918	10.18
0.1561	8.0	800	0.2006	99.9251	99.9188	99.9251	99.9251	10.17
0.1561	9.0	900	0.1628	99.9251	99.9188	99.9251	99.9251	10.23
0.1538	10.0	1000	0.1881	99.9251	99.9188	99.9251	99.9251	10.17

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1