9e8e91a54447b85ca8f0a21ddea5dcb9

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

• Loss: 1.6498
• Data Size: 1.0
• Epoch Runtime: 352.9511
• Bleu: 16.4904

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	4.7552	0	29.1799	1.5029
No log	1	1407	1.9891	0.0078	32.7496	8.5334
No log	2	2814	1.8371	0.0156	35.0267	9.9730
0.048	3	4221	1.7347	0.0312	40.3059	11.2355
1.7056	4	5628	1.6588	0.0625	52.0337	17.0455
1.5969	5	7035	1.5649	0.125	72.5298	13.7998
1.429	6	8442	1.4782	0.25	112.1122	15.1856
1.3394	7	9849	1.3897	0.5	189.4133	16.6813
1.1634	8.0	11256	1.3392	1.0	351.6894	15.2476
0.9433	9.0	12663	1.3446	1.0	349.8489	15.0066
0.8003	10.0	14070	1.4263	1.0	349.5650	18.8023
0.5976	11.0	15477	1.5266	1.0	348.1628	15.1415
0.4663	12.0	16884	1.6498	1.0	352.9511	16.4904

Framework versions

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