nllb-mulgi

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: 0.6669

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.1832	1.0	5381	1.0899
0.9676	2.0	10762	0.9221
0.8225	3.0	16143	0.8334
0.7285	4.0	21524	0.7831
0.6536	5.0	26905	0.7447
0.5835	6.0	32286	0.7133
0.5209	7.0	37667	0.6919
0.4773	8.0	43048	0.6752
0.4316	9.0	48429	0.6662
0.3868	10.0	53810	0.6621
0.3783	11.0	59191	0.6568
0.3298	12.0	64572	0.6571
0.3217	13.0	69953	0.6558
0.2898	14.0	75334	0.6569
0.2803	15.0	80715	0.6606
0.2513	16.0	86096	0.6597
0.2418	17.0	91477	0.6644
0.2302	18.0	96858	0.6647
0.2235	19.0	102239	0.6665
0.2259	20.0	107620	0.6669

Framework versions

• Transformers 5.7.0
• Pytorch 2.6.0+cu126
• Datasets 4.8.5
• Tokenizers 0.22.2