finetuning_summarization

This model is a fine-tuned version of cahya/bert2bert-indonesian-summarization on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6759
• Rouge1: 0.8455
• Rouge2: 0.742
• Rougel: 0.8486
• Rougelsum: 0.8475
• Gen Len: 23.7368

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: 16
• eval_batch_size: 16
• 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	5	1.3699	0.8443	0.7258	0.8426	0.8435	25.8421
No log	2.0	10	1.0257	0.8282	0.7115	0.8293	0.8275	25.0
No log	3.0	15	0.7871	0.8384	0.7277	0.8397	0.8396	24.3158
No log	4.0	20	0.7078	0.8339	0.7318	0.8358	0.8348	23.4211
No log	5.0	25	0.6994	0.843	0.7396	0.8451	0.845	24.0
No log	6.0	30	0.6832	0.8445	0.7413	0.8419	0.842	23.4737
No log	7.0	35	0.6768	0.8429	0.742	0.8451	0.8448	23.6842
No log	8.0	40	0.6736	0.843	0.7396	0.8451	0.845	23.6842
No log	9.0	45	0.6750	0.843	0.7396	0.8451	0.845	23.6842
No log	10.0	50	0.6759	0.8455	0.742	0.8486	0.8475	23.7368

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2