1522cb02f8be48124311ba5f4a35309b

This model is a fine-tuned version of facebook/mbart-large-50-many-to-one-mmt on the Helsinki-NLP/opus_books [fr-ru] dataset. It achieves the following results on the evaluation set:

• Loss: 2.7075
• Data Size: 1.0
• Epoch Runtime: 55.2417
• Bleu: 5.4574

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Bleu
No log	0	0	6.3344	0	4.8810	0.5513
No log	1	204	5.6099	0.0078	5.7054	0.7932
No log	2	408	4.8306	0.0156	6.7366	1.0641
No log	3	612	4.1149	0.0312	9.1358	1.6992
No log	4	816	3.7139	0.0625	10.7286	2.3232
No log	5	1020	3.3973	0.125	13.1248	2.9385
0.2954	6	1224	3.0912	0.25	20.1530	3.5158
2.7741	7	1428	2.7846	0.5	32.2649	4.0305
2.2659	8.0	1632	2.5336	1.0	56.3917	4.8119
1.7095	9.0	1836	2.4783	1.0	54.4150	5.3225
1.2996	10.0	2040	2.4913	1.0	54.1450	5.4219
0.9326	11.0	2244	2.5382	1.0	54.7570	5.3958
0.647	12.0	2448	2.6286	1.0	53.4898	5.5752
0.4789	13.0	2652	2.7075	1.0	55.2417	5.4574

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.2.0
• Tokenizers 0.22.1