2cc4670158bd1ff104f6fa33a700e2ce

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

• Loss: 2.6717
• Data Size: 1.0
• Epoch Runtime: 182.0175
• Bleu: 6.2878

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	8.0420	0	15.7912	0.2500
No log	1	688	4.5315	0.0078	17.3488	0.6961
No log	2	1376	3.6073	0.0156	19.9333	1.5752
No log	3	2064	3.1689	0.0312	23.1490	2.2464
0.133	4	2752	2.9034	0.0625	28.5437	2.8428
0.2425	5	3440	2.7440	0.125	38.9532	3.3638
2.7148	6	4128	2.5909	0.25	59.8023	4.0315
2.4841	7	4816	2.4455	0.5	100.7491	4.6082
2.2484	8.0	5504	2.3165	1.0	186.1785	5.6393
1.9551	9.0	6192	2.2906	1.0	180.6179	5.6007
1.7105	10.0	6880	2.3365	1.0	181.6472	6.5945
1.4714	11.0	7568	2.4109	1.0	180.4393	7.1280
1.2655	12.0	8256	2.5328	1.0	181.3078	6.0857
1.0555	13.0	8944	2.6717	1.0	182.0175	6.2878

Framework versions

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