bart-base-japanese-tobyoki-pairwise

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

• Loss: 3.5252
• Rouge1: 11.814
• Rouge2: 1.7965
• Rougel: 8.0177
• Rougelsum: 9.7342
• Gen Len: 50.4446

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: 5e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
0.2994	1.0	4332	2.7883	11.1611	1.7768	7.5158	9.6222	55.0633
0.1513	2.0	8664	3.1286	13.7182	2.311	9.1726	11.5058	57.3528
0.0778	3.0	12996	3.3238	12.1173	1.88	8.1156	10.1187	48.7089
0.056	4.0	17328	3.4032	11.9555	2.0536	8.2185	10.0656	50.7373
0.0364	5.0	21660	3.5252	11.814	1.7965	8.0177	9.7342	50.4446

Framework versions

• Transformers 4.30.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.4
• Tokenizers 0.13.3