nllb_muse_aligned

This model is a fine-tuned version of Koubra-Gaby/nllb_muse_aligned on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.4506
• Bleu: 7.3485
• Rouge: 0.3402
• Meteor: 0.2219
• Chrf++: 25.5126

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: 64
• seed: 42
• 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: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Rouge	Meteor	Chrf++
4.4085	1.0438	500	3.5328	0.8739	0.1473	0.0672	9.1212
3.3811	2.0877	1000	3.0764	1.3954	0.1691	0.0779	11.1657
2.9566	3.1315	1500	2.7647	1.7621	0.1833	0.0893	12.8896
2.6531	4.1754	2000	2.5463	2.2216	0.2029	0.103	14.4948
2.4248	5.2192	2500	2.3851	2.6681	0.2253	0.1193	16.1982
2.2414	6.2630	3000	2.2577	3.2792	0.2401	0.1328	17.5515
2.1042	7.3069	3500	2.1569	3.6682	0.2535	0.1432	18.7198
1.9741	8.3507	4000	2.0719	4.1136	0.2656	0.153	19.6689
1.8706	9.3946	4500	2.0013	4.479	0.2707	0.1597	20.3466
1.7847	10.4384	5000	1.9412	4.7734	0.2771	0.1668	20.9034
1.6995	11.4823	5500	1.8906	4.99	0.2839	0.1718	21.4169
1.6277	12.5261	6000	1.8428	5.22	0.2897	0.1766	21.7939
1.5579	13.5699	6500	1.8024	5.434	0.2931	0.1809	22.2236
1.4999	14.6138	7000	1.7700	5.5704	0.2998	0.1851	22.564
1.4503	15.6576	7500	1.7363	5.8033	0.3043	0.1891	22.9528
1.3972	16.7015	8000	1.7071	5.9462	0.3053	0.1911	23.0934
1.3542	17.7453	8500	1.6815	6.0391	0.308	0.1931	23.342
1.3062	18.7891	9000	1.6614	6.2428	0.3117	0.1966	23.596
1.2737	19.8330	9500	1.6403	6.2823	0.3126	0.197	23.7064
1.2325	20.8768	10000	1.6233	6.4767	0.3145	0.1993	23.9034
1.1958	21.9207	10500	1.6034	6.5277	0.3176	0.2011	24.0746
1.1600	22.9645	11000	1.5897	6.7215	0.3195	0.2036	24.2943
1.1317	24.0084	11500	1.5735	6.7212	0.3202	0.204	24.3597
1.1022	25.0522	12000	1.5674	6.7104	0.3235	0.2058	24.5052
1.0737	26.0960	12500	1.5573	6.8112	0.3257	0.2077	24.6729
1.0471	27.1399	13000	1.5475	6.89	0.3258	0.208	24.746
1.0247	28.1837	13500	1.5374	6.9748	0.3251	0.21	24.7983
0.9970	29.2276	14000	1.5259	7.0124	0.3265	0.2101	24.9217
0.9757	30.2714	14500	1.5275	7.0323	0.3277	0.211	24.9938
0.9552	31.3152	15000	1.5185	7.1292	0.3286	0.2124	25.0567
0.9367	32.3591	15500	1.5086	7.1978	0.3302	0.2136	25.1342
0.9139	33.4029	16000	1.5059	7.2436	0.3308	0.2146	25.178
0.8976	34.4468	16500	1.5004	7.2487	0.3323	0.2153	25.2457
0.8753	35.4906	17000	1.4960	7.1641	0.3334	0.2154	25.2409
0.8594	36.5344	17500	1.4937	7.2567	0.3333	0.2157	25.3283
0.8434	37.5783	18000	1.4912	7.3168	0.3343	0.2168	25.4111
0.8279	38.6221	18500	1.4884	7.3246	0.333	0.2168	25.3448
0.8131	39.6660	19000	1.4879	7.3889	0.3344	0.2175	25.4399
0.7982	40.7098	19500	1.4868	7.4074	0.3349	0.2184	25.4948
0.7824	41.7537	20000	1.4875	7.5111	0.3386	0.2219	25.6273
0.7739	42.7975	20500	1.4790	7.4232	0.3376	0.2205	25.5501
0.7570	43.8413	21000	1.4868	7.5504	0.3396	0.2225	25.7074
0.7436	44.8852	21500	1.4821	7.4953	0.3393	0.2213	25.6226
0.7339	45.9290	22000	1.4805	7.4989	0.3381	0.2212	25.6702

Framework versions

• Transformers 5.21.0
• Pytorch 2.11.0+cu130
• Datasets 4.8.5
• Tokenizers 0.23.3