29004e70c0b068445756e946fd41c0e8

This model is a fine-tuned version of google/mt5-small on the Helsinki-NLP/opus_books [fr-pt] dataset. It achieves the following results on the evaluation set:

• Loss: 2.3858
• Data Size: 1.0
• Epoch Runtime: 6.0970
• Bleu: 4.0242

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	27.5438	0	1.1526	0.0058
No log	1	31	28.4233	0.0078	1.2516	0.0063
No log	2	62	28.0428	0.0156	1.7271	0.0045
No log	3	93	26.5597	0.0312	2.1117	0.0059
No log	4	124	24.4093	0.0625	2.3954	0.0041
No log	5	155	22.2079	0.125	2.7253	0.0082
No log	6	186	17.6759	0.25	3.1398	0.0094
3.5054	7	217	13.5341	0.5	4.4290	0.0117
3.5054	8.0	248	9.8188	1.0	6.9570	0.0154
10.928	9.0	279	7.8998	1.0	6.9664	0.0167
10.4524	10.0	310	6.6437	1.0	6.9542	0.0186
10.4524	11.0	341	5.4714	1.0	7.4791	0.0186
7.6418	12.0	372	4.6749	1.0	7.6182	0.0317
6.0055	13.0	403	3.7281	1.0	7.4343	0.0887
6.0055	14.0	434	3.3707	1.0	8.3537	0.3530
4.9541	15.0	465	3.1714	1.0	5.9927	1.1423
4.9541	16.0	496	3.0677	1.0	5.9569	1.2514
4.4308	17.0	527	2.9753	1.0	5.8289	1.5034
4.1102	18.0	558	2.9014	1.0	6.2157	1.7000
4.1102	19.0	589	2.8467	1.0	6.3563	1.8398
3.9069	20.0	620	2.7960	1.0	6.8002	2.2300
3.7412	21.0	651	2.7565	1.0	7.2603	2.3792
3.7412	22.0	682	2.7301	1.0	7.5230	2.4908
3.6364	23.0	713	2.6889	1.0	7.9714	2.7077
3.6364	24.0	744	2.6600	1.0	9.1440	2.8903
3.498	25.0	775	2.6445	1.0	8.9545	2.9441
3.4366	26.0	806	2.6197	1.0	8.8483	2.9626
3.4366	27.0	837	2.5967	1.0	8.5830	3.0582
3.3506	28.0	868	2.5836	1.0	6.0944	3.1140
3.3506	29.0	899	2.5682	1.0	6.0129	3.1509
3.2866	30.0	930	2.5502	1.0	5.9571	3.1536
3.2124	31.0	961	2.5412	1.0	6.0268	3.2192
3.2124	32.0	992	2.5242	1.0	6.0170	3.2674
3.1606	33.0	1023	2.5124	1.0	6.0481	3.2754
3.1155	34.0	1054	2.4978	1.0	5.8964	3.2844
3.1155	35.0	1085	2.4921	1.0	5.9531	3.4115
3.0601	36.0	1116	2.4796	1.0	6.3192	3.5079
3.0601	37.0	1147	2.4673	1.0	6.8573	3.5407
2.998	38.0	1178	2.4623	1.0	7.3572	3.6119
2.9555	39.0	1209	2.4581	1.0	8.2948	3.6769
2.9555	40.0	1240	2.4465	1.0	8.2879	3.6685
2.9102	41.0	1271	2.4395	1.0	8.4370	3.7489
2.8699	42.0	1302	2.4366	1.0	8.7327	3.7991
2.8699	43.0	1333	2.4264	1.0	5.8849	3.7835
2.8366	44.0	1364	2.4188	1.0	5.8377	3.8390
2.8366	45.0	1395	2.4157	1.0	5.9253	3.9188
2.7911	46.0	1426	2.4078	1.0	6.1507	3.9007
2.764	47.0	1457	2.4002	1.0	5.9784	3.9577
2.764	48.0	1488	2.3913	1.0	5.9478	3.9526
2.7201	49.0	1519	2.3885	1.0	5.9027	3.9700
2.6912	50.0	1550	2.3858	1.0	6.0970	4.0242

Framework versions

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