bart-base-japanese-tobyoki-pairwise-wo_space

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: 3.2694
• Rouge1: 18.7407
• Rouge2: 3.1211
• Rougel: 10.9379
• Rougelsum: 15.8203
• Gen Len: 95.0245

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
0.7833	1.0	2025	2.5751	16.3343	1.2888	11.0128	15.2802	65.3776
0.3308	2.0	4050	2.9423	17.9514	2.8091	10.9133	15.4068	91.6503
0.2302	3.0	6075	3.0625	16.1453	3.0026	10.272	14.0716	77.1993
0.1576	4.0	8100	3.2308	17.8409	2.9937	10.8765	15.6203	88.3986
0.1055	5.0	10125	3.2694	18.7407	3.1211	10.9379	15.8203	95.0245

Framework versions

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