2c4e74418f0fb555a85ce5bd8d7f88eb

This model is a fine-tuned version of google/mt5-large on the Helsinki-NLP/opus_books [de-nl] dataset. It achieves the following results on the evaluation set:

• Loss: 1.8539
• Data Size: 1.0
• Epoch Runtime: 157.9288
• Bleu: 9.0135

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	23.7303	0	12.4678	0.0208
No log	1	390	24.1355	0.0078	14.1786	0.0184
No log	2	780	24.3028	0.0156	16.6559	0.0166
No log	3	1170	17.3621	0.0312	21.3792	0.0179
No log	4	1560	14.1326	0.0625	26.5135	0.0209
1.1012	5	1950	9.8915	0.125	36.6540	0.0183
1.7496	6	2340	4.3676	0.25	55.1910	0.1699
3.1138	7	2730	2.3169	0.5	93.5042	5.3114
2.5365	8.0	3120	2.0728	1.0	169.0967	6.5409
2.3024	9.0	3510	1.9927	1.0	162.8727	7.0852
2.1691	10.0	3900	1.9378	1.0	160.1259	7.6590
2.0854	11.0	4290	1.8989	1.0	162.4107	7.9075
1.989	12.0	4680	1.8734	1.0	163.0841	8.1366
1.9149	13.0	5070	1.8588	1.0	160.3492	8.2804
1.8378	14.0	5460	1.8429	1.0	160.8582	8.5066
1.7263	15.0	5850	1.8366	1.0	160.3609	8.6285
1.6765	16.0	6240	1.8364	1.0	162.3152	8.7134
1.6365	17.0	6630	1.8284	1.0	161.5530	8.7564
1.5684	18.0	7020	1.8330	1.0	160.1011	8.7800
1.5244	19.0	7410	1.8346	1.0	159.7138	8.9650
1.4834	20.0	7800	1.8405	1.0	158.3638	8.9365
1.4328	21.0	8190	1.8539	1.0	157.9288	9.0135

Framework versions

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