0ee5a78fd2220aa7891c0075b50125fc

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

• Loss: 0.6485
• Data Size: 1.0
• Epoch Runtime: 28.3177
• Bleu: 13.4729

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	5.7206	0	2.6305	0.1139
No log	1	447	3.1931	0.0078	3.0803	0.0079
0.0589	2	894	2.5179	0.0156	3.1478	0.0922
0.0588	3	1341	2.0219	0.0312	3.5702	0.7033
0.0823	4	1788	1.7522	0.0625	4.2978	1.5766
0.1313	5	2235	1.5345	0.125	5.8991	2.4306
1.5614	6	2682	1.3262	0.25	9.0927	4.0661
1.2508	7	3129	1.1098	0.5	15.3096	5.5438
0.9889	8.0	3576	0.8984	1.0	28.0396	7.8445
0.8658	9.0	4023	0.7966	1.0	27.8408	9.1197
0.7701	10.0	4470	0.7366	1.0	27.6390	10.1494
0.6991	11.0	4917	0.6996	1.0	28.4780	10.8819
0.6441	12.0	5364	0.6771	1.0	27.7038	11.3733
0.5868	13.0	5811	0.6531	1.0	29.2186	11.8695
0.5479	14.0	6258	0.6410	1.0	28.7030	12.3612
0.519	15.0	6705	0.6351	1.0	27.7515	12.5152
0.4724	16.0	7152	0.6347	1.0	28.2174	12.6411
0.4515	17.0	7599	0.6260	1.0	28.1993	12.7953
0.4262	18.0	8046	0.6206	1.0	28.0471	13.1307
0.4082	19.0	8493	0.6346	1.0	27.2427	13.2125
0.3806	20.0	8940	0.6277	1.0	28.7176	13.2173
0.3685	21.0	9387	0.6454	1.0	28.9584	13.1375
0.3311	22.0	9834	0.6485	1.0	28.3177	13.4729

Framework versions

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