a4b022bf2c72dcfa49cc964a7758fd65

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

• Loss: 2.7259
• Data Size: 1.0
• Epoch Runtime: 21.8422
• Bleu: 1.7634

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	16.7344	0	2.1982	0.0183
No log	1	70	17.9500	0.0078	2.8194	0.0225
No log	2	140	18.4235	0.0156	2.8075	0.0144
No log	3	210	17.6207	0.0312	3.6709	0.0206
No log	4	280	15.1332	0.0625	4.6320	0.0227
No log	5	350	11.9948	0.125	5.8862	0.0249
No log	6	420	10.6293	0.25	8.5793	0.0225
2.1061	7	490	7.9887	0.5	11.8413	0.0298
7.9092	8.0	560	4.0808	1.0	18.4734	0.0213
5.1198	9.0	630	3.1978	1.0	18.8138	0.6244
3.9301	10.0	700	3.0048	1.0	18.1908	1.0309
3.7242	11.0	770	2.9259	1.0	19.0205	1.1184
3.5688	12.0	840	2.8641	1.0	20.3791	1.1812
3.3971	13.0	910	2.8347	1.0	17.0713	1.1238
3.3314	14.0	980	2.8092	1.0	17.8308	1.2414
3.2193	15.0	1050	2.7889	1.0	18.3994	1.3361
3.1528	16.0	1120	2.7650	1.0	17.6708	1.3915
3.0915	17.0	1190	2.7547	1.0	17.0606	1.4049
2.9862	18.0	1260	2.7448	1.0	18.1235	1.4754
2.9561	19.0	1330	2.7311	1.0	19.0352	1.3886
2.8636	20.0	1400	2.7293	1.0	19.2752	1.5498
2.8108	21.0	1470	2.7184	1.0	20.2734	1.5849
2.8107	22.0	1540	2.7178	1.0	21.2743	1.6642
2.7031	23.0	1610	2.7173	1.0	17.1539	1.6211
2.6608	24.0	1680	2.7170	1.0	17.9669	1.6293
2.6397	25.0	1750	2.7168	1.0	17.7668	1.7854
2.5663	26.0	1820	2.7083	1.0	18.0221	1.7601
2.5616	27.0	1890	2.7258	1.0	19.2925	1.7630
2.4963	28.0	1960	2.7285	1.0	18.0288	1.7219
2.4589	29.0	2030	2.7360	1.0	21.3086	1.7601
2.4366	30.0	2100	2.7259	1.0	21.8422	1.7634

Framework versions

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