AraT5-base-finetune-ar-xlsum

This model is a fine-tuned version of UBC-NLP/AraT5-base on the xlsum dataset. It achieves the following results on the evaluation set:

• Loss: 4.4714
• Rouge-1: 29.55
• Rouge-2: 12.63
• Rouge-l: 25.8
• Gen Len: 18.76
• Bertscore: 73.3

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• num_epochs: 10
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge-1	Rouge-2	Rouge-l	Gen Len	Bertscore
11.9753	1.0	293	7.0887	11.93	2.56	10.93	17.19	63.85
6.7818	2.0	586	5.7712	19.94	6.34	17.65	18.64	69.0
5.9434	3.0	879	5.1083	23.51	8.56	20.66	18.88	70.78
5.451	4.0	1172	4.8538	25.84	10.05	22.63	18.42	72.04
5.1643	5.0	1465	4.6910	27.23	11.13	23.83	18.78	72.45
4.9693	6.0	1758	4.5950	28.42	11.71	24.82	18.74	72.94
4.8308	7.0	2051	4.5323	28.95	12.19	25.3	18.74	73.13
4.7284	8.0	2344	4.4956	29.19	12.37	25.53	18.76	73.18
4.653	9.0	2637	4.4757	29.44	12.48	25.63	18.78	73.23
4.606	10.0	2930	4.4714	29.55	12.63	25.8	18.76	73.3

Framework versions

• Transformers 4.19.4
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1