AraT5-base-title-generation-finetune-ar-xlsum

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

• Loss: 4.2837
• Rouge-1: 32.46
• Rouge-2: 15.15
• Rouge-l: 28.38
• Gen Len: 18.48
• Bertscore: 74.24

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
5.815	1.0	293	4.7437	27.05	10.49	23.56	18.03	72.56
5.0818	2.0	586	4.5004	28.92	11.97	25.09	18.61	73.08
4.7855	3.0	879	4.3910	29.66	12.57	25.79	18.58	73.3
4.588	4.0	1172	4.3469	30.22	13.05	26.36	18.59	73.61
4.4388	5.0	1465	4.3226	30.88	13.81	27.01	18.65	73.78
4.3162	6.0	1758	4.2990	30.9	13.6	26.92	18.68	73.78
4.2178	7.0	2051	4.2869	31.35	14.01	27.41	18.57	73.96
4.1387	8.0	2344	4.2794	31.28	13.98	27.34	18.6	73.87
4.0787	9.0	2637	4.2806	31.45	14.17	27.46	18.66	73.97
4.0371	10.0	2930	4.2837	31.55	14.19	27.52	18.65	74.0

Framework versions

• Transformers 4.20.0
• Pytorch 1.11.0+cu113
• Datasets 2.3.2
• Tokenizers 0.12.1