76b9d373e2cd50b6ae3ec8def3dd22e3

This model is a fine-tuned version of facebook/mbart-large-en-ro on the Helsinki-NLP/opus_books [fr-it] dataset. It achieves the following results on the evaluation set:

• Loss: 2.7165
• Data Size: 1.0
• Epoch Runtime: 100.2904
• Bleu: 6.8036

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	7.9193	0	9.2246	0.2274
No log	1	367	5.3304	0.0078	10.0641	0.2995
No log	2	734	4.0833	0.0156	11.3887	1.2460
No log	3	1101	3.4734	0.0312	14.2217	2.6989
No log	4	1468	3.1209	0.0625	17.9417	4.4296
0.1865	5	1835	2.8730	0.125	24.5648	4.5408
2.8813	6	2202	2.6470	0.25	36.2180	5.6733
2.5507	7	2569	2.4786	0.5	56.3885	6.6551
2.264	8.0	2936	2.3703	1.0	104.1389	7.3225
1.9665	9.0	3303	2.3590	1.0	101.2593	6.5271
1.6641	10.0	3670	2.3936	1.0	101.9876	6.7849
1.421	11.0	4037	2.4993	1.0	101.5019	6.8285
1.2345	12.0	4404	2.5964	1.0	101.3380	6.5625
1.0568	13.0	4771	2.7165	1.0	100.2904	6.8036

Framework versions

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