70271a268c3e123382014df2a1971831

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

• Loss: 5.8639
• Data Size: 1.0
• Epoch Runtime: 22.1241
• Bleu: 0.3164

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	13.6352	0	2.4635	0.1990
No log	1	70	11.3085	0.0078	2.9932	0.2009
No log	2	140	11.0147	0.0156	4.5462	0.2207
No log	3	210	9.2760	0.0312	5.5653	0.0571
No log	4	280	8.3625	0.0625	7.5766	0.1924
No log	5	350	7.5944	0.125	9.6046	0.1633
No log	6	420	6.6443	0.25	11.2748	0.2749
1.126	7	490	5.7002	0.5	14.6275	0.6001
5.2593	8.0	560	5.0453	1.0	23.8047	0.7805
4.7225	9.0	630	4.7522	1.0	23.1385	0.9869
4.0016	10.0	700	4.5100	1.0	23.6459	1.0757
4.2758	11.0	770	11.2109	1.0	21.8179	0.0
11.9106	12.0	840	6.7697	1.0	22.6605	0.0077
6.5717	13.0	910	6.4409	1.0	23.5267	0.1066
6.162	14.0	980	5.8639	1.0	22.1241	0.3164

Framework versions

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