3d60c57c3f3c5f84a08a67aca192dceb

This model is a fine-tuned version of google/umt5-base on the Helsinki-NLP/opus_books [fi-fr] dataset. It achieves the following results on the evaluation set:

• Loss: 2.4253
• Data Size: 1.0
• Epoch Runtime: 23.7459
• Bleu: 4.5379

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	12.4930	0	2.4647	0.0252
No log	1	88	12.3720	0.0078	2.7830	0.0249
No log	2	176	12.3626	0.0156	4.1122	0.0317
No log	3	264	12.1881	0.0312	5.8347	0.0239
No log	4	352	11.3909	0.0625	7.4028	0.0239
No log	5	440	9.8327	0.125	9.6713	0.0376
1.1949	6	528	8.3388	0.25	12.5681	0.0799
4.2766	7	616	7.0456	0.5	15.9200	0.1479
5.9558	8.0	704	3.9423	1.0	24.7735	1.9463
4.9277	9.0	792	3.0999	1.0	22.7905	4.5165
4.1288	10.0	880	2.7956	1.0	22.5833	2.5227
3.7293	11.0	968	2.7002	1.0	22.8304	2.8830
3.4868	12.0	1056	2.6330	1.0	22.6622	3.2056
3.3872	13.0	1144	2.5942	1.0	24.2142	3.3695
3.2691	14.0	1232	2.5500	1.0	24.7831	3.4857
3.1419	15.0	1320	2.5322	1.0	22.9104	3.4446
3.0114	16.0	1408	2.5091	1.0	23.4552	3.8765
2.9133	17.0	1496	2.4738	1.0	25.0048	3.9387
2.8438	18.0	1584	2.4587	1.0	24.7820	3.9834
2.7913	19.0	1672	2.4573	1.0	22.6510	4.0498
2.697	20.0	1760	2.4375	1.0	22.6254	4.1260
2.6554	21.0	1848	2.4304	1.0	22.7013	4.1289
2.6171	22.0	1936	2.4352	1.0	24.4042	4.2317
2.5397	23.0	2024	2.4222	1.0	23.3454	4.2221
2.4962	24.0	2112	2.4267	1.0	23.9381	4.2985
2.4463	25.0	2200	2.4151	1.0	23.3682	4.4046
2.4017	26.0	2288	2.4168	1.0	23.7511	4.4193
2.3341	27.0	2376	2.4161	1.0	26.1075	4.4194
2.3059	28.0	2464	2.4077	1.0	22.7957	4.4997
2.2576	29.0	2552	2.4237	1.0	22.7708	4.5188
2.2294	30.0	2640	2.4263	1.0	22.9204	4.4313
2.1692	31.0	2728	2.4236	1.0	24.2041	4.4810
2.1514	32.0	2816	2.4253	1.0	23.7459	4.5379

Framework versions

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