Bamalingua-4

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: 0.2322
• Bleu: 26.1291

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
0.332	1.0	3160	0.3005	17.7092
0.2854	2.0	6320	0.2758	20.3331
0.2492	3.0	9480	0.2624	21.5687
0.2248	4.0	12640	0.2531	23.2174
0.2122	5.0	15800	0.2464	23.7911
0.1978	6.0	18960	0.2414	24.5054
0.182	7.0	22120	0.2376	24.7621
0.1804	8.0	25280	0.2353	25.5139
0.168	9.0	28440	0.2343	25.7965
0.1609	10.0	31600	0.2328	25.8732
0.1583	11.0	34760	0.2323	26.1820
0.1603	12.0	37920	0.2322	26.1291

Framework versions

• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.4.2
• Tokenizers 0.22.1