a3bdaa46b3e502d2a06cd428b5d3ae1e

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

• Loss: 2.0601
• Data Size: 1.0
• Epoch Runtime: 46.9453
• Bleu: 5.7971

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	18.6095	0	4.5211	0.0136
No log	1	204	17.3576	0.0078	4.7122	0.0142
No log	2	408	16.3045	0.0156	5.5679	0.0169
No log	3	612	14.8927	0.0312	6.2537	0.0159
No log	4	816	11.1767	0.0625	8.6712	0.0145
No log	5	1020	7.9354	0.125	11.9308	0.0252
0.9848	6	1224	5.3314	0.25	15.4465	0.0287
5.14	7	1428	2.9558	0.5	26.3808	0.6859
3.6072	8.0	1632	2.5230	1.0	47.0342	2.3668
3.2278	9.0	1836	2.3959	1.0	44.3959	2.7395
3.0851	10.0	2040	2.3240	1.0	44.8476	3.0738
2.9153	11.0	2244	2.2805	1.0	45.1252	3.2457
2.8097	12.0	2448	2.2381	1.0	44.5890	3.4658
2.6992	13.0	2652	2.1911	1.0	44.6900	3.7137
2.5957	14.0	2856	2.1635	1.0	45.4663	3.8334
2.5173	15.0	3060	2.1350	1.0	45.7136	3.9924
2.4215	16.0	3264	2.1216	1.0	46.7592	4.1316
2.3519	17.0	3468	2.1107	1.0	45.0083	4.2839
2.3378	18.0	3672	2.0925	1.0	46.0805	4.4017
2.2692	19.0	3876	2.0784	1.0	44.8357	4.4737
2.2006	20.0	4080	2.0757	1.0	46.2836	4.8063
2.1543	21.0	4284	2.0690	1.0	47.1755	4.8514
2.0956	22.0	4488	2.0558	1.0	47.1702	5.0655
2.0687	23.0	4692	2.0561	1.0	46.9332	4.9966
2.0073	24.0	4896	2.0529	1.0	45.2441	4.9802
1.9752	25.0	5100	2.0567	1.0	46.7553	5.0820
1.927	26.0	5304	2.0518	1.0	46.3650	5.2447
1.8755	27.0	5508	2.0517	1.0	46.3734	5.1145
1.8388	28.0	5712	2.0515	1.0	46.9446	5.4780
1.779	29.0	5916	2.0525	1.0	45.3471	5.4312
1.7595	30.0	6120	2.0545	1.0	44.8351	5.8170
1.7471	31.0	6324	2.0574	1.0	46.3363	5.6903
1.7126	32.0	6528	2.0601	1.0	46.9453	5.7971

Framework versions

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