d2c05477e598bc8e977a6fde50e93c65

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

• Loss: 2.1704
• Data Size: 1.0
• Epoch Runtime: 24.0078
• Bleu: 3.5016

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	8.3352	0	2.4477	0.0385
No log	1	367	6.7973	0.0078	2.8787	0.0346
No log	2	734	6.0998	0.0156	3.0607	0.1110
No log	3	1101	5.5015	0.0312	3.4085	0.1200
No log	4	1468	4.9414	0.0625	3.8055	0.1246
0.2744	5	1835	4.3896	0.125	5.1987	0.2520
4.2833	6	2202	3.8692	0.25	8.0355	0.5011
3.6834	7	2569	3.3766	0.5	13.1599	0.9881
3.1167	8.0	2936	2.9024	1.0	24.9190	1.6340
2.77	9.0	3303	2.6728	1.0	23.9793	2.0650
2.5528	10.0	3670	2.5337	1.0	23.9125	2.3801
2.3945	11.0	4037	2.4404	1.0	23.3820	2.6259
2.2611	12.0	4404	2.3530	1.0	22.7753	2.7747
2.199	13.0	4771	2.3055	1.0	24.0031	2.9132
2.0573	14.0	5138	2.2429	1.0	22.8630	3.0419
2.0159	15.0	5505	2.2100	1.0	23.6841	3.1650
1.9134	16.0	5872	2.1989	1.0	23.0399	3.2144
1.8726	17.0	6239	2.1886	1.0	24.4095	3.2784
1.8023	18.0	6606	2.1634	1.0	23.5569	3.2946
1.7615	19.0	6973	2.1575	1.0	24.4966	3.3313
1.693	20.0	7340	2.1550	1.0	23.9728	3.3850
1.6236	21.0	7707	2.1339	1.0	24.1947	3.4222
1.5797	22.0	8074	2.1411	1.0	23.6790	3.4273
1.5059	23.0	8441	2.1489	1.0	23.5070	3.4422
1.4669	24.0	8808	2.1586	1.0	24.5366	3.4802
1.4461	25.0	9175	2.1704	1.0	24.0078	3.5016

Framework versions

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