mt5-base-finetuned-persian-finetuned-persian-arabic

This model is a fine-tuned version of ahmeddbahaa/mt5-base-finetuned-persian on the xlsum dataset. It achieves the following results on the evaluation set:

• Loss: 3.3234
• Rouge-1: 22.96
• Rouge-2: 10.27
• Rouge-l: 20.95
• Gen Len: 19.0
• Bertscore: 71.59

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge-1	Rouge-2	Rouge-l	Gen Len	Bertscore
4.2754	1.0	1172	3.5717	19.26	7.26	17.48	19.0	70.49
3.7388	2.0	2344	3.4291	19.71	7.88	17.94	19.0	70.64
3.541	3.0	3516	3.3653	21.18	8.84	19.35	19.0	71.05
3.4113	4.0	4688	3.3306	21.54	9.11	19.65	19.0	71.19
3.3256	5.0	5860	3.3234	21.69	9.22	19.81	19.0	71.31

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu113
• Datasets 2.1.0
• Tokenizers 0.12.1