53206cd9f8a5c89a5c063d2df26e6587

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

• Loss: 2.4075
• Data Size: 1.0
• Epoch Runtime: 56.8618
• Bleu: 6.1217

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	26.6082	0	5.3460	0.0037
No log	1	390	24.8507	0.0078	6.9534	0.0039
No log	2	780	21.8076	0.0156	6.6067	0.0060
No log	3	1170	18.9334	0.0312	8.1142	0.0033
No log	4	1560	14.8386	0.0625	9.2352	0.0070
1.0192	5	1950	9.1086	0.125	12.6142	0.0117
1.537	6	2340	5.9527	0.25	19.1072	0.0181
5.3284	7	2730	3.7119	0.5	31.2303	1.1817
4.326	8.0	3120	3.3122	1.0	56.5618	2.1861
3.9947	9.0	3510	3.1591	1.0	56.2062	2.6803
3.8043	10.0	3900	3.0513	1.0	56.8248	3.0562
3.6641	11.0	4290	2.9802	1.0	57.5752	3.4920
3.5628	12.0	4680	2.9200	1.0	57.0053	3.7333
3.5145	13.0	5070	2.8800	1.0	56.2891	3.9421
3.4306	14.0	5460	2.8305	1.0	56.1577	4.1149
3.3156	15.0	5850	2.7955	1.0	56.5411	4.2152
3.2754	16.0	6240	2.7686	1.0	54.6214	4.4542
3.253	17.0	6630	2.7444	1.0	55.0890	4.4858
3.1731	18.0	7020	2.7225	1.0	56.6047	4.5586
3.148	19.0	7410	2.6913	1.0	55.6237	4.6820
3.1067	20.0	7800	2.6790	1.0	55.5254	4.7541
3.0622	21.0	8190	2.6594	1.0	56.0049	4.8190
3.0174	22.0	8580	2.6329	1.0	56.9300	4.8789
3.0019	23.0	8970	2.6250	1.0	55.3320	4.9717
2.9484	24.0	9360	2.6063	1.0	55.1256	5.0273
2.9187	25.0	9750	2.5972	1.0	56.3781	5.0626
2.9095	26.0	10140	2.5853	1.0	56.1055	5.1179
2.8524	27.0	10530	2.5713	1.0	56.4429	5.2100
2.8516	28.0	10920	2.5601	1.0	55.4443	5.2673
2.8325	29.0	11310	2.5430	1.0	56.2263	5.3145
2.8002	30.0	11700	2.5319	1.0	56.3833	5.3782
2.735	31.0	12090	2.5281	1.0	56.4733	5.4505
2.7312	32.0	12480	2.5184	1.0	56.4822	5.4686
2.7051	33.0	12870	2.5092	1.0	56.9100	5.5074
2.6724	34.0	13260	2.5067	1.0	55.9261	5.5596
2.674	35.0	13650	2.4875	1.0	55.5195	5.6071
2.6742	36.0	14040	2.4842	1.0	56.7434	5.6734
2.6162	37.0	14430	2.4774	1.0	56.5541	5.7158
2.581	38.0	14820	2.4693	1.0	55.7454	5.7728
2.5671	39.0	15210	2.4665	1.0	56.4713	5.7896
2.5664	40.0	15600	2.4549	1.0	55.3665	5.8258
2.5588	41.0	15990	2.4533	1.0	56.0915	5.8644
2.545	42.0	16380	2.4522	1.0	56.1284	5.8635
2.5216	43.0	16770	2.4378	1.0	55.3636	5.9064
2.5006	44.0	17160	2.4335	1.0	58.8562	5.9244
2.4887	45.0	17550	2.4291	1.0	58.5788	5.9757
2.4745	46.0	17940	2.4264	1.0	58.2341	6.0034
2.4107	47.0	18330	2.4197	1.0	59.1548	6.0582
2.4954	48.0	18720	2.4161	1.0	58.2660	6.0710
2.4066	49.0	19110	2.4099	1.0	58.7295	6.0844
2.4026	50.0	19500	2.4075	1.0	56.8618	6.1217

Framework versions

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