nllb-200-distilled-600M-mt-finetuned-zindi-dyu-to-fr

This model is a fine-tuned version of facebook/nllb-200-distilled-600M on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.2584
• Bleu: 6.4075

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: 64
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
3.1707	0.1575	20	2.7356	4.8084
2.9074	0.3150	40	2.5883	5.0141
2.7168	0.4724	60	2.4902	5.5785
2.6912	0.6299	80	2.4154	5.7743
2.6062	0.7874	100	2.3742	6.0010
2.5794	0.9449	120	2.3480	6.1354
2.4634	1.1024	140	2.3314	5.9899
2.5055	1.2598	160	2.3167	6.1080
2.5062	1.4173	180	2.3032	6.3784
2.4771	1.5748	200	2.2944	6.4510
2.4284	1.7323	220	2.2854	6.2883
2.4423	1.8898	240	2.2783	6.5036
2.3202	2.0472	260	2.2730	6.4039
2.3855	2.2047	280	2.2701	6.2921
2.4292	2.3622	300	2.2658	6.3025
2.3678	2.5197	320	2.2626	6.2881
2.4158	2.6772	340	2.2600	6.3684
2.351	2.8346	360	2.2588	6.2852
2.3755	2.9921	380	2.2584	6.2819

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1