b4143f6ca2b670480e5e0bf5f1a591c7

This model is a fine-tuned version of google/mt5-base on the Helsinki-NLP/opus_books [fr-no] dataset. It achieves the following results on the evaluation set:

• Loss: 2.4553
• Data Size: 1.0
• Epoch Runtime: 21.4284
• Bleu: 4.5235

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	15.3497	0	2.3715	0.0022
No log	1	86	15.5050	0.0078	2.4274	0.0018
No log	2	172	15.0724	0.0156	3.2516	0.0019
No log	3	258	14.3781	0.0312	4.1122	0.0011
No log	4	344	13.0770	0.0625	4.9695	0.0022
0.8365	5	430	12.7477	0.125	6.0761	0.0028
3.4542	6	516	9.9971	0.25	8.3372	0.0026
3.8575	7	602	7.1582	0.5	12.3786	0.0048
4.2106	8.0	688	3.3526	1.0	21.2431	0.6846
3.9921	9.0	774	2.8347	1.0	20.0207	2.4839
3.5189	10.0	860	2.7125	1.0	20.0402	2.9441
3.3456	11.0	946	2.6347	1.0	19.7216	3.2853
3.1552	12.0	1032	2.5922	1.0	19.8099	3.4452
3.0832	13.0	1118	2.5679	1.0	21.9515	3.6581
2.9532	14.0	1204	2.5390	1.0	20.0691	3.6913
2.8685	15.0	1290	2.5244	1.0	20.0207	4.0225
2.8127	16.0	1376	2.5045	1.0	20.2384	4.0064
2.7368	17.0	1462	2.4830	1.0	19.7876	4.0810
2.673	18.0	1548	2.4794	1.0	20.5133	4.0975
2.5977	19.0	1634	2.4854	1.0	20.1231	4.1771
2.5595	20.0	1720	2.4650	1.0	19.4532	4.2873
2.5266	21.0	1806	2.4530	1.0	20.2203	4.2258
2.4756	22.0	1892	2.4611	1.0	20.8453	4.3009
2.4141	23.0	1978	2.4540	1.0	22.1911	4.3980
2.4037	24.0	2064	2.4610	1.0	20.8924	4.3724
2.3271	25.0	2150	2.4553	1.0	21.4284	4.5235

Framework versions

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