ec9a614c892a5ce57e8eb6439ad3bd51

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

• Loss: 2.8520
• Data Size: 1.0
• Epoch Runtime: 8.0758
• Bleu: 5.1945

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	16.1748	0	1.1777	0.4226
No log	1	31	16.1153	0.0078	1.3125	0.4585
No log	2	62	15.9533	0.0156	1.6327	0.4890
No log	3	93	15.8908	0.0312	1.9235	0.5583
No log	4	124	15.8601	0.0625	2.0207	0.3142
No log	5	155	15.5561	0.125	2.6232	0.3385
No log	6	186	15.6011	0.25	3.4505	0.3442
2.9313	7	217	14.4831	0.5	4.8066	0.3555
2.9313	8.0	248	11.6977	1.0	7.6944	0.3809
10.718	9.0	279	9.6328	1.0	7.8142	0.3768
12.0868	10.0	310	8.1565	1.0	8.3106	0.3889
12.0868	11.0	341	7.1047	1.0	8.9109	0.2825
9.558	12.0	372	6.3523	1.0	8.9924	0.4299
8.0857	13.0	403	5.4300	1.0	9.0836	0.9043
8.0857	14.0	434	4.7490	1.0	6.3490	2.0336
7.0196	15.0	465	4.4147	1.0	6.6850	3.5048
7.0196	16.0	496	4.2541	1.0	6.8621	4.3999
6.2394	17.0	527	4.1485	1.0	7.2159	4.8407
5.7914	18.0	558	4.0515	1.0	8.4663	5.1538
5.7914	19.0	589	3.9218	1.0	7.5790	5.6354
5.4274	20.0	620	3.8342	1.0	7.6054	3.8640
5.1882	21.0	651	3.7428	1.0	7.5607	2.1286
5.1882	22.0	682	3.6675	1.0	7.4714	1.7915
4.9689	23.0	713	3.5918	1.0	7.5897	1.8241
4.9689	24.0	744	3.5217	1.0	7.9090	1.8391
4.7617	25.0	775	3.4577	1.0	8.3472	1.2336
4.6285	26.0	806	3.3979	1.0	8.1950	1.2248
4.6285	27.0	837	3.3432	1.0	8.5740	1.2389
4.4538	28.0	868	3.2944	1.0	6.4040	1.3340
4.4538	29.0	899	3.2475	1.0	7.1924	1.3961
4.3293	30.0	930	3.2026	1.0	7.1671	1.4277
4.2152	31.0	961	3.1681	1.0	7.2722	1.7774
4.2152	32.0	992	3.1311	1.0	7.1802	3.3872
4.1074	33.0	1023	3.1071	1.0	7.5470	6.7424
4.0027	34.0	1054	3.0775	1.0	7.5095	5.6128
4.0027	35.0	1085	3.0529	1.0	7.5647	4.8327
3.9314	36.0	1116	3.0317	1.0	7.8623	4.6476
3.9314	37.0	1147	3.0122	1.0	7.8229	4.6880
3.8486	38.0	1178	3.0019	1.0	7.6763	4.6790
3.7769	39.0	1209	2.9852	1.0	8.2100	4.7554
3.7769	40.0	1240	2.9691	1.0	9.0335	4.8626
3.7066	41.0	1271	2.9450	1.0	8.9545	5.0563
3.6342	42.0	1302	2.9380	1.0	6.8437	5.0005
3.6342	43.0	1333	2.9260	1.0	6.9673	5.0035
3.5889	44.0	1364	2.9099	1.0	6.9993	5.0697
3.5889	45.0	1395	2.8987	1.0	7.4151	4.9567
3.5232	46.0	1426	2.8914	1.0	7.8934	4.9850
3.4865	47.0	1457	2.8773	1.0	8.4549	5.0317
3.4865	48.0	1488	2.8704	1.0	8.1827	5.1013
3.4057	49.0	1519	2.8602	1.0	7.9880	5.1254
3.3826	50.0	1550	2.8520	1.0	8.0758	5.1945

Framework versions

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