43e6d5ec1cc998b32b9a054ed5c18b52

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

• Loss: 2.0862
• Data Size: 1.0
• Epoch Runtime: 321.0671
• Bleu: 9.8662

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	2.3366	0	27.3916	9.0018
No log	1	1286	1.7666	0.0078	29.9859	9.0212
0.0362	2	2572	1.6984	0.0156	33.0030	9.5952
0.046	3	3858	1.6608	0.0312	38.0487	9.9181
0.0698	4	5144	1.6314	0.0625	47.7796	11.4613
1.5565	5	6430	1.6009	0.125	66.5578	14.0338
1.5151	6	7716	1.5729	0.25	102.8124	11.0096
1.4306	7	9002	1.5413	0.5	175.2207	13.5821
1.3268	8.0	10288	1.5176	1.0	323.5635	12.5303
1.0681	9.0	11574	1.5698	1.0	322.8706	10.8120
0.8311	10.0	12860	1.6999	1.0	322.8238	10.3587
0.6287	11.0	14146	1.9043	1.0	323.6919	10.0241
0.5118	12.0	15432	2.0862	1.0	321.0671	9.8662

Framework versions

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