opus-mt-tc-big-en-ar-finetuned-franco-to-arabic-3

This model is a fine-tuned version of Helsinki-NLP/opus-mt-tc-big-en-ar on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.0027
• Bleu: 9.3993
• Gen Len: 14.8889

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
No log	1.0	21	8.1233	0.0	217.5833
No log	2.0	42	7.3522	0.1587	139.6111
No log	3.0	63	6.7128	0.2113	63.9722
No log	4.0	84	6.2283	0.2611	25.7222
No log	5.0	105	5.7868	0.2545	27.3333
No log	6.0	126	5.4596	0.4153	25.75
No log	7.0	147	5.1424	0.8436	12.6944
No log	8.0	168	4.8195	0.8365	13.3889
No log	9.0	189	4.6181	1.2564	12.3889
No log	10.0	210	4.3480	1.7124	12.9722
No log	11.0	231	4.1206	3.2404	13.4167
No log	12.0	252	3.8982	2.4371	14.4167
No log	13.0	273	3.7679	3.9697	14.2778
No log	14.0	294	3.6359	4.7946	13.8889
No log	15.0	315	3.4940	5.9909	14.75
No log	16.0	336	3.3826	8.193	14.5278
No log	17.0	357	3.3084	7.5299	14.3056
No log	18.0	378	3.2302	8.5741	14.5833
No log	19.0	399	3.1751	5.9148	14.8611
No log	20.0	420	3.1242	10.389	14.3889
No log	21.0	441	3.0618	10.5285	14.6944
No log	22.0	462	3.0439	11.3953	14.6667
No log	23.0	483	3.0161	10.9885	14.8889
4.3311	24.0	504	3.0184	9.6143	14.7778
4.3311	25.0	525	3.0027	9.3993	14.8889

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3