mt5-base-finetuned-ar-sp

This model is a fine-tuned version of google/mt5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.2772
• Rouge-1: 23.01
• Rouge-2: 10.41
• Rouge-l: 20.94
• Gen Len: 19.0
• Bertscore: 71.56

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
4.1968	1.0	1352	3.5142	18.69	6.73	16.97	19.0	70.3
3.6932	2.0	2704	3.3799	20.67	8.38	18.75	19.0	70.82
3.5058	3.0	4056	3.3184	20.97	8.58	19.08	19.0	71.08
3.3832	4.0	5408	3.2851	21.59	8.94	19.63	19.0	71.28
3.2994	5.0	6760	3.2772	21.84	9.23	19.85	19.0	71.34

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1