t5-finetuned-ar-to-arsl_test_bleu

This model is a fine-tuned version of PRAli22/arat5-arabic-dialects-translation on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3564
• Bleu1: 0.9326
• Bleu2: 0.9010
• Bleu3: 0.7310
• Bleu4: 0.5945
• Overall Bleu: 0.7773

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu1	Bleu2	Bleu3	Bleu4	Overall Bleu
No log	1.0	69	0.3152	0.9226	0.8885	0.7160	0.5780	0.7632
No log	2.0	138	0.2849	0.9285	0.8941	0.7237	0.5864	0.7704
No log	3.0	207	0.3064	0.9311	0.8992	0.7285	0.5913	0.7750
No log	4.0	276	0.2908	0.9326	0.8984	0.7290	0.5914	0.7753
No log	5.0	345	0.3007	0.9354	0.9034	0.7352	0.5975	0.7805
No log	6.0	414	0.3043	0.9315	0.8991	0.7311	0.5940	0.7766
No log	7.0	483	0.3124	0.9311	0.8981	0.7297	0.5941	0.7760
0.1232	8.0	552	0.3217	0.9302	0.8968	0.7254	0.5889	0.7726
0.1232	9.0	621	0.3201	0.9316	0.8989	0.7293	0.5946	0.7763
0.1232	10.0	690	0.3370	0.9317	0.8974	0.7280	0.5932	0.7752
0.1232	11.0	759	0.3458	0.9335	0.9013	0.7317	0.5954	0.7781
0.1232	12.0	828	0.3446	0.9324	0.8993	0.7291	0.5924	0.7758
0.1232	13.0	897	0.3516	0.9337	0.9017	0.7304	0.5938	0.7774
0.1232	14.0	966	0.3570	0.9331	0.9014	0.7309	0.5942	0.7774
0.0507	15.0	1035	0.3564	0.9326	0.9010	0.7310	0.5945	0.7773

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2