lamso-en-nllb

This model is a fine-tuned version of facebook/nllb-200-distilled-600M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.1536
• Bleu: 5.4661

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• 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
• lr_scheduler_warmup_steps: 50
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu
5.5569	1.0	40	4.6347	0.0949
4.583	2.0	80	3.8912	0.2283
4.0616	3.0	120	3.5883	0.2882
3.698	4.0	160	3.4173	0.6949
3.5286	5.0	200	3.2860	1.0174
3.3239	6.0	240	3.2013	1.0944
3.2044	7.0	280	3.1336	1.1160
3.0838	8.0	320	3.1032	1.5282
2.982	9.0	360	3.0618	2.1756
2.84	10.0	400	3.0452	2.2410
2.7283	11.0	440	3.0317	2.4307
2.7494	12.0	480	3.0313	2.6356
2.5419	13.0	520	3.0275	2.3977
2.6218	14.0	560	3.0123	2.8629
2.5425	15.0	600	3.0041	2.8604
2.461	16.0	640	3.0067	3.3949
2.4507	17.0	680	3.0097	3.1292
2.2282	18.0	720	3.0183	3.5313
2.4489	19.0	760	3.0126	3.5240
2.144	20.0	800	3.0304	3.7798
2.1395	21.0	840	3.0248	4.0427
2.066	22.0	880	3.0380	3.9976
2.1631	23.0	920	3.0704	4.0978
2.038	24.0	960	3.0579	4.1935
2.0262	25.0	1000	3.0652	4.1297
2.0747	26.0	1040	3.0674	4.4140
1.8927	27.0	1080	3.0789	4.5061
1.8739	28.0	1120	3.0898	4.4716
1.8864	29.0	1160	3.0870	4.7618
1.9223	30.0	1200	3.0986	4.6710
1.9868	31.0	1240	3.1137	4.9288
1.9936	32.0	1280	3.1169	4.6553
1.9889	33.0	1320	3.1330	4.5970
1.906	34.0	1360	3.1260	4.7602
1.8468	35.0	1400	3.1294	4.6674
1.8345	36.0	1440	3.1246	5.0111
1.8759	37.0	1480	3.1325	4.5826
1.688	38.0	1520	3.1386	4.8248
1.9076	39.0	1560	3.1418	5.0801
1.681	40.0	1600	3.1414	5.1977
1.6575	41.0	1640	3.1469	5.1851
1.7737	42.0	1680	3.1456	4.9597
1.7835	43.0	1720	3.1512	5.2460
1.796	44.0	1760	3.1401	5.2646
1.8359	45.0	1800	3.1444	5.2608
1.7384	46.0	1840	3.1464	5.3784
1.724	47.0	1880	3.1515	5.3768
1.7753	48.0	1920	3.1536	5.4661
1.5805	49.0	1960	3.1526	5.3439
1.7253	50.0	2000	3.1536	5.2808

Framework versions

• Transformers 4.57.2
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1