t5-finetuned-ar-to-arsl_test

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: 0.3309
• Bleu1: 0.9310
• Bleu2: 0.8974
• Bleu3: 0.7219
• Bleu4: 0.5884

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
No log	1.0	59	0.4660	0.8440	0.7803	0.5955	0.4626
No log	2.0	118	0.3223	0.8966	0.8500	0.6689	0.5321
No log	2.99	177	0.3004	0.9170	0.8791	0.7022	0.5644
No log	3.99	236	0.2925	0.9205	0.8834	0.7071	0.5703
No log	4.99	295	0.3099	0.9223	0.8859	0.7090	0.5716
No log	5.99	354	0.2879	0.9244	0.8892	0.7125	0.5768
No log	6.99	413	0.2971	0.9280	0.8936	0.7176	0.5824
No log	8.0	473	0.2986	0.9254	0.8899	0.7136	0.5800
0.3874	9.0	532	0.3128	0.9293	0.8952	0.7204	0.5874
0.3874	10.0	591	0.3166	0.9316	0.8992	0.7242	0.5907
0.3874	10.99	650	0.3270	0.9303	0.8964	0.7214	0.5861
0.3874	11.99	709	0.3290	0.9304	0.8961	0.7223	0.5883
0.3874	12.99	768	0.3326	0.9296	0.8957	0.7216	0.5880
0.3874	13.99	827	0.3309	0.9294	0.8959	0.7208	0.5870
0.3874	14.97	885	0.3309	0.9310	0.8974	0.7219	0.5884

Framework versions

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