1ad297f0d273d3bdafaf30ab0c28a5ef

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

• Loss: 3.0512
• Data Size: 1.0
• Epoch Runtime: 19.9753
• Bleu: 5.3748

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	8.9400	0	2.5971	0.1210
No log	1	58	5.5454	0.0078	2.5452	0.7510
No log	2	116	4.6370	0.0156	3.0369	1.4175
No log	3	174	3.9067	0.0312	4.8981	2.7275
No log	4	232	3.4844	0.0625	6.4374	5.8324
No log	5	290	3.2030	0.125	8.4214	4.2232
0.325	6	348	2.9968	0.25	10.2308	4.5295
0.4037	7	406	2.7888	0.5	13.6448	5.6627
1.817	8.0	464	2.6626	1.0	19.8449	6.1915
2.0847	9.0	522	2.6202	1.0	20.5314	5.6680
1.7523	10.0	580	2.6882	1.0	20.6363	5.9930
1.5088	11.0	638	2.8161	1.0	19.0945	5.4628
1.2642	12.0	696	2.9188	1.0	19.5026	5.3680
0.8684	13.0	754	3.0512	1.0	19.9753	5.3748

Framework versions

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