5d95478d467aad242abe03c24af6cdb7

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

• Loss: 2.5742
• Data Size: 1.0
• Epoch Runtime: 54.7279
• Bleu: 15.4780

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	6.8208	0	4.8496	0.2383
No log	1	204	3.7001	0.0078	5.4791	1.3875
No log	2	408	3.0300	0.0156	6.7090	2.3953
No log	3	612	2.8504	0.0312	8.8952	3.0131
No log	4	816	2.7000	0.0625	10.3860	3.6681
No log	5	1020	2.5404	0.125	13.2221	4.9321
0.2132	6	1224	2.4052	0.25	19.8468	5.3369
2.1593	7	1428	2.2549	0.5	31.7315	6.6811
1.8387	8.0	1632	2.1302	1.0	55.8651	7.4763
1.3908	9.0	1836	2.1612	1.0	54.1049	8.6313
1.0455	10.0	2040	2.2723	1.0	53.7569	14.8606
0.7441	11.0	2244	2.4132	1.0	54.2865	13.2564
0.496	12.0	2448	2.5742	1.0	54.7279	15.4780

Framework versions

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