99191bcaa9c115bebb03ad6623d6c935

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

• Loss: 3.0938
• Data Size: 1.0
• Epoch Runtime: 27.1049
• Bleu: 9.2717

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	5.8966	0	2.4315	0.4536
No log	1	88	4.5552	0.0078	2.9276	0.6250
No log	2	176	3.6024	0.0156	4.5407	1.2883
No log	3	264	3.1083	0.0312	6.6403	1.7134
No log	4	352	2.8836	0.0625	8.1161	2.1738
No log	5	440	2.7178	0.125	10.2417	2.8955
0.2237	6	528	2.5746	0.25	12.7966	3.7449
0.8222	7	616	2.4424	0.5	16.9490	4.7494
2.0104	8.0	704	2.3406	1.0	28.6189	6.3345
1.5874	9.0	792	2.3934	1.0	28.7683	9.3491
1.166	10.0	880	2.5395	1.0	25.2322	9.5132
0.8543	11.0	968	2.8161	1.0	26.2926	9.8069
0.5847	12.0	1056	3.0938	1.0	27.1049	9.2717

Framework versions

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