043ad4c0673b7b805e55ffcf7c7facd2

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

• Loss: 2.8468
• Data Size: 1.0
• Epoch Runtime: 9.3855
• Bleu: 6.9328

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	14.5531	0	1.2274	0.3691
No log	1	33	14.4282	0.0078	1.5663	0.3678
No log	2	66	14.3336	0.0156	1.8326	0.4480
No log	3	99	13.9828	0.0312	2.4607	0.5592
0.9865	4	132	13.7348	0.0625	2.7346	0.5176
0.9865	5	165	13.3052	0.125	2.8951	0.3762
0.9865	6	198	12.5736	0.25	3.6002	0.4177
3.2784	7	231	11.5112	0.5	4.8297	0.2903
10.1975	8.0	264	9.4828	1.0	7.8565	0.3259
10.1975	9.0	297	8.0425	1.0	7.6858	0.3395
11.542	10.0	330	7.0185	1.0	7.7051	0.2910
9.2216	11.0	363	6.1231	1.0	7.8345	0.5347
9.2216	12.0	396	5.2366	1.0	8.2612	1.5508
7.7374	13.0	429	4.7289	1.0	8.1415	2.8071
6.8788	14.0	462	4.4008	1.0	6.4158	3.1602
6.8788	15.0	495	4.1853	1.0	6.9048	4.8227
6.2517	16.0	528	4.0325	1.0	6.9100	5.8487
5.8105	17.0	561	3.9273	1.0	7.3793	6.5909
5.8105	18.0	594	3.8128	1.0	7.4730	8.0173
5.4397	19.0	627	3.7417	1.0	7.8123	6.4436
5.1515	20.0	660	3.6720	1.0	7.6148	3.4242
5.1515	21.0	693	3.6022	1.0	8.1808	2.8380
4.9315	22.0	726	3.5335	1.0	8.4593	2.5799
4.7371	23.0	759	3.4811	1.0	8.5840	2.6716
4.7371	24.0	792	3.4170	1.0	8.9305	2.9427
4.5438	25.0	825	3.3588	1.0	9.2967	2.9134
4.3807	26.0	858	3.3087	1.0	6.7780	1.8310
4.3807	27.0	891	3.2600	1.0	7.2956	1.8348
4.2275	28.0	924	3.2261	1.0	7.1792	1.8874
4.0842	29.0	957	3.1805	1.0	7.3200	1.9388
4.0842	30.0	990	3.1462	1.0	7.6242	1.9511
3.979	31.0	1023	3.1241	1.0	7.7700	4.1224
3.8889	32.0	1056	3.0762	1.0	7.7625	11.7766
3.8889	33.0	1089	3.0536	1.0	7.7024	11.3380
3.7607	34.0	1122	3.0227	1.0	8.1078	6.2495
3.7125	35.0	1155	3.0149	1.0	7.7447	6.1134
3.7125	36.0	1188	2.9815	1.0	8.0411	6.0309
3.6541	37.0	1221	2.9789	1.0	8.1344	6.0923
3.5645	38.0	1254	2.9486	1.0	8.4616	6.2092
3.5645	39.0	1287	2.9382	1.0	8.4584	6.3690
3.5157	40.0	1320	2.9208	1.0	6.9229	6.5477
3.435	41.0	1353	2.9159	1.0	7.1035	6.4658
3.435	42.0	1386	2.9058	1.0	7.1838	6.6646
3.3653	43.0	1419	2.8924	1.0	7.5417	6.8209
3.3331	44.0	1452	2.8816	1.0	7.8022	6.7550
3.3331	45.0	1485	2.8825	1.0	7.9223	6.8957
3.266	46.0	1518	2.8642	1.0	8.1227	6.9220
3.2248	47.0	1551	2.8609	1.0	8.2907	6.9719
3.2248	48.0	1584	2.8468	1.0	8.7553	6.9140
3.1796	49.0	1617	2.8478	1.0	8.8618	6.9574
3.1305	50.0	1650	2.8468	1.0	9.3855	6.9328

Framework versions

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