70ffa625b72fcb7d9bac25974d4a500b

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

• Loss: 2.3317
• Data Size: 1.0
• Epoch Runtime: 47.6352
• Bleu: 3.2167

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	7.9944	0	2.9340	0.0099
No log	1	70	6.4267	0.0078	3.8022	0.0143
No log	2	140	4.4921	0.0156	10.8559	0.0789
No log	3	210	4.0037	0.0312	16.8179	0.1364
No log	4	280	3.6246	0.0625	24.5894	0.2065
No log	5	350	3.1522	0.125	25.7154	0.2285
No log	6	420	2.5981	0.25	24.5447	1.3129
0.5735	7	490	2.3739	0.5	32.3747	2.1926
2.5888	8.0	560	2.2219	1.0	51.5136	2.3441
2.3169	9.0	630	2.1865	1.0	43.3882	2.5375
2.0089	10.0	700	2.1677	1.0	47.4489	2.7884
1.8212	11.0	770	2.1955	1.0	42.8757	3.0258
1.7045	12.0	840	2.2293	1.0	47.2441	3.4181
1.4757	13.0	910	2.2720	1.0	43.3185	3.4087
1.3741	14.0	980	2.3317	1.0	47.6352	3.2167

Framework versions

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