d0578ba508daa23e602199ff56a2cb4b

This model is a fine-tuned version of facebook/mbart-large-cc25 on the Helsinki-NLP/opus_books [de-es] dataset. It achieves the following results on the evaluation set:

• Loss: 2.5965
• Data Size: 1.0
• Epoch Runtime: 180.7257
• Bleu: 6.1554

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	9.8137	0	16.0250	0.0855
No log	1	688	4.0155	0.0078	17.3887	1.6471
No log	2	1376	3.5569	0.0156	20.0431	2.8302
No log	3	2064	3.1735	0.0312	23.2252	3.1177
0.1317	4	2752	2.9158	0.0625	28.3684	5.2247
0.2356	5	3440	2.6268	0.125	38.7669	3.5663
2.5928	6	4128	2.4739	0.25	58.9114	3.9224
2.367	7	4816	2.3424	0.5	100.2041	4.6788
2.1966	8.0	5504	2.2657	1.0	181.8426	5.0631
1.861	9.0	6192	2.2880	1.0	179.1325	6.7576
1.6528	10.0	6880	2.3269	1.0	180.5646	5.9854
1.3728	11.0	7568	2.4052	1.0	180.4532	5.7398
1.1397	12.0	8256	2.5965	1.0	180.7257	6.1554

Framework versions

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