5c480ca36dd43f52b75f74195eb73b0e

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

• Loss: 2.4095
• Data Size: 1.0
• Epoch Runtime: 34.1895
• Bleu: 2.5139

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	20.7821	0	3.1376	0.0146
No log	1	70	20.0721	0.0078	3.3194	0.0128
No log	2	140	17.1994	0.0156	5.4450	0.0139
No log	3	210	11.5444	0.0312	8.4575	0.0160
No log	4	280	6.7808	0.0625	10.9733	0.0144
No log	5	350	5.5241	0.125	12.9772	0.0208
No log	6	420	4.1001	0.25	17.4450	0.0621
1.2817	7	490	3.3514	0.5	22.6773	0.1693
3.8438	8.0	560	2.8316	1.0	37.0722	0.2368
3.3647	9.0	630	2.5680	1.0	34.3399	1.5915
2.9383	10.0	700	2.4821	1.0	34.7818	1.9042
2.7904	11.0	770	2.4330	1.0	36.0166	1.8460
2.6826	12.0	840	2.4230	1.0	34.3227	1.9706
2.5249	13.0	910	2.3987	1.0	35.3792	1.9738
2.4542	14.0	980	2.3968	1.0	34.7106	2.1376
2.3639	15.0	1050	2.3903	1.0	34.3147	2.2314
2.2788	16.0	1120	2.3847	1.0	35.1708	2.3091
2.2334	17.0	1190	2.3854	1.0	35.1582	2.4028
2.1489	18.0	1260	2.4079	1.0	34.2471	2.5268
2.0943	19.0	1330	2.4064	1.0	34.4802	2.5671
2.003	20.0	1400	2.4095	1.0	34.1895	2.5139

Framework versions

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