ea498f2f8800ae82b2b80e3f3e8bf7e9

This model is a fine-tuned version of google/mt5-small on the Helsinki-NLP/opus_books [fr-no] dataset. It achieves the following results on the evaluation set:

• Loss: 2.8706
• Data Size: 1.0
• Epoch Runtime: 13.7158
• Bleu: 2.8681

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	24.6813	0	1.7986	0.0009
No log	1	86	24.7378	0.0078	2.8486	0.0008
No log	2	172	24.3812	0.0156	2.3142	0.0008
No log	3	258	23.0589	0.0312	2.7868	0.0014
No log	4	344	20.8463	0.0625	3.5438	0.0015
1.3681	5	430	18.1595	0.125	4.3812	0.0027
5.2509	6	516	15.6510	0.25	6.1068	0.0020
5.5813	7	602	10.3997	0.5	8.8942	0.0033
6.4297	8.0	688	5.4761	1.0	15.3246	0.0122
6.5951	9.0	774	4.0052	1.0	13.4675	0.2429
5.0986	10.0	860	3.6742	1.0	14.4898	0.4658
4.841	11.0	946	3.5062	1.0	14.6042	0.6311
4.4852	12.0	1032	3.4093	1.0	14.1982	0.7627
4.2868	13.0	1118	3.3364	1.0	14.5429	1.0343
4.138	14.0	1204	3.2795	1.0	14.1576	1.2543
4.0571	15.0	1290	3.2347	1.0	14.0612	1.3775
3.9666	16.0	1376	3.2016	1.0	13.3305	1.4498
3.8844	17.0	1462	3.1669	1.0	13.7788	1.5649
3.8281	18.0	1548	3.1439	1.0	14.5092	1.6304
3.7532	19.0	1634	3.1191	1.0	14.4961	1.6749
3.7084	20.0	1720	3.0959	1.0	14.8120	1.8000
3.6705	21.0	1806	3.0784	1.0	14.5898	1.8761
3.6522	22.0	1892	3.0642	1.0	14.6089	1.8872
3.5636	23.0	1978	3.0461	1.0	13.4842	1.9836
3.551	24.0	2064	3.0356	1.0	13.7628	2.0694
3.4916	25.0	2150	3.0229	1.0	14.1647	2.0876
3.4554	26.0	2236	3.0042	1.0	14.4404	2.1515
3.4224	27.0	2322	2.9986	1.0	15.4571	2.1357
3.3942	28.0	2408	2.9844	1.0	15.4711	2.1600
3.3785	29.0	2494	2.9769	1.0	15.4662	2.2534
3.3361	30.0	2580	2.9689	1.0	15.7904	2.2643
3.3061	31.0	2666	2.9625	1.0	14.0855	2.2744
3.2623	32.0	2752	2.9541	1.0	13.9531	2.3525
3.2314	33.0	2838	2.9447	1.0	13.9750	2.3598
3.2346	34.0	2924	2.9440	1.0	14.0985	2.4240
3.2265	35.0	3010	2.9302	1.0	14.2094	2.4658
3.1866	36.0	3096	2.9258	1.0	13.8554	2.4391
3.1598	37.0	3182	2.9187	1.0	14.1753	2.4766
3.1336	38.0	3268	2.9156	1.0	14.1022	2.4855
3.0847	39.0	3354	2.9069	1.0	13.5672	2.5021
3.0866	40.0	3440	2.9001	1.0	13.4121	2.5222
3.0738	41.0	3526	2.8992	1.0	13.9182	2.5807
3.0578	42.0	3612	2.8924	1.0	14.3426	2.6003
3.0272	43.0	3698	2.8916	1.0	15.0226	2.6529
3.0087	44.0	3784	2.8856	1.0	15.5647	2.7099
2.9882	45.0	3870	2.8847	1.0	15.5668	2.7470
2.9678	46.0	3956	2.8827	1.0	15.8990	2.7260
2.9595	47.0	4042	2.8777	1.0	13.5686	2.7856
2.9079	48.0	4128	2.8710	1.0	13.6765	2.8488
2.9165	49.0	4214	2.8716	1.0	13.7148	2.8691
2.8845	50.0	4300	2.8706	1.0	13.7158	2.8681

Framework versions

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