1ea3e77587a769d98bb55cbf63fbb152

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

• Loss: 2.9909
• Data Size: 1.0
• Epoch Runtime: 22.6085
• Bleu: 3.0276

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	14.1670	0	2.2825	0.0400
No log	1	83	13.9621	0.0078	2.7123	0.0444
No log	2	166	13.8706	0.0156	2.9007	0.0483
No log	3	249	14.0058	0.0312	3.9104	0.0487
0.578	4	332	13.5798	0.0625	4.6265	0.0465
0.578	5	415	13.2520	0.125	6.1050	0.0427
0.578	6	498	12.4261	0.25	8.1859	0.0465
3.0514	7	581	11.4465	0.5	12.8473	0.0810
11.0976	8.0	664	7.6243	1.0	21.9064	0.1305
8.3957	9.0	747	4.5427	1.0	21.2292	0.9970
5.6798	10.0	830	3.9434	1.0	22.1189	0.8173
4.9084	11.0	913	3.5901	1.0	22.2857	1.1404
4.6296	12.0	996	3.4043	1.0	21.2481	1.5943
4.2624	13.0	1079	3.3149	1.0	23.0058	1.8751
4.0565	14.0	1162	3.2505	1.0	23.5358	2.0493
3.9554	15.0	1245	3.1943	1.0	24.1467	2.1809
3.7762	16.0	1328	3.1451	1.0	21.6642	2.4197
3.6523	17.0	1411	3.1145	1.0	22.2990	2.4832
3.6047	18.0	1494	3.0874	1.0	23.6836	2.5480
3.4779	19.0	1577	3.0682	1.0	24.3783	2.6775
3.3998	20.0	1660	3.0451	1.0	22.1055	2.6308
3.3042	21.0	1743	3.0210	1.0	23.4489	2.6541
3.2199	22.0	1826	3.0148	1.0	24.0244	2.7412
3.1597	23.0	1909	3.0017	1.0	22.3515	2.7230
3.1211	24.0	1992	2.9940	1.0	22.4391	2.7134
3.0261	25.0	2075	2.9855	1.0	22.3969	2.8133
2.9905	26.0	2158	2.9877	1.0	22.3190	2.8429
2.9268	27.0	2241	2.9913	1.0	23.0228	2.8606
2.8712	28.0	2324	2.9972	1.0	22.1533	2.8658
2.8487	29.0	2407	2.9827	1.0	21.6873	2.8885
2.8161	30.0	2490	2.9852	1.0	23.1436	2.8758
2.7398	31.0	2573	2.9970	1.0	23.1308	2.9793
2.7125	32.0	2656	2.9911	1.0	24.0320	3.0432
2.6314	33.0	2739	2.9909	1.0	22.6085	3.0276

Framework versions

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