led-risalah_data_v17_3

This model is a fine-tuned version of silmi224/finetune-led-35000 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.6551
• Rouge1: 25.33
• Rouge2: 12.4758
• Rougel: 18.3801
• Rougelsum: 24.0275

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
2.9826	1.0	20	2.5250	11.8736	3.4553	8.0701	10.4233
2.5516	2.0	40	2.2001	15.7664	5.0213	10.8555	14.1975
2.2334	3.0	60	2.0424	17.0425	6.006	10.956	15.2795
1.9577	4.0	80	1.9305	19.1792	7.6754	12.651	17.7519
1.8602	5.0	100	1.8351	22.4846	8.3095	14.0022	20.587
1.702	6.0	120	1.7809	21.9395	8.5042	14.9427	20.3436
1.6525	7.0	140	1.7286	23.7825	10.9231	15.9319	22.0902
1.5285	8.0	160	1.6839	24.1286	11.2382	16.7057	22.3731
1.4623	9.0	180	1.6644	23.8767	12.3834	17.5761	22.6869
1.4175	10.0	200	1.6551	25.33	12.4758	18.3801	24.0275

Framework versions

• Transformers 4.41.2
• Pytorch 2.1.2
• Datasets 2.19.2
• Tokenizers 0.19.1