7b1586f73092dfe8c4a95f3ea86eb15b

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

• Loss: 2.0831
• Data Size: 1.0
• Epoch Runtime: 219.3822
• Bleu: 11.6308

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	6.6359	0	18.5116	0.6542
No log	1	872	4.3977	0.0078	21.1232	3.4745
No log	2	1744	3.3645	0.0156	23.2054	4.0781
0.0654	3	2616	2.6953	0.0312	27.6742	4.8906
0.1688	4	3488	2.0340	0.0625	34.4863	9.3080
2.1619	5	4360	10.9287	0.125	47.3261	0.0
5.7218	6	5232	8.5979	0.25	73.4600	0.0007
1.8686	7	6104	1.8531	0.5	121.3464	17.0301
1.6253	8.0	6976	1.7161	1.0	221.0925	18.2478
1.3286	9.0	7848	1.6849	1.0	218.2240	18.1615
1.0815	10.0	8720	1.7036	1.0	218.7692	17.3426
0.9058	11.0	9592	1.8086	1.0	219.1696	18.4044
0.6767	12.0	10464	1.9659	1.0	222.0382	12.6996
0.523	13.0	11336	2.0831	1.0	219.3822	11.6308

Framework versions

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