4f6b0c53057ae1c793ce15d7bc7b775f

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

• Loss: 1.3503
• Data Size: 1.0
• Epoch Runtime: 181.0558
• Bleu: 10.6815

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	23.0086	0	13.8254	0.0177
No log	1	437	24.2370	0.0078	15.5508	0.0181
No log	2	874	22.4692	0.0156	18.1560	0.0179
No log	3	1311	20.1321	0.0312	21.8500	0.0203
No log	4	1748	12.3472	0.0625	27.5813	0.0235
6.758	5	2185	2.8360	0.125	38.6283	0.4308
3.1699	6	2622	1.9156	0.25	59.6038	2.9574
2.229	7	3059	1.6252	0.5	101.4152	6.5201
1.8745	8.0	3496	1.4850	1.0	184.1214	8.2572
1.7347	9.0	3933	1.4291	1.0	184.4869	9.5415
1.587	10.0	4370	1.3908	1.0	186.9955	9.8106
1.5171	11.0	4807	1.3615	1.0	183.2168	10.2186
1.424	12.0	5244	1.3433	1.0	182.3575	10.3990
1.3361	13.0	5681	1.3335	1.0	184.1796	10.3927
1.2673	14.0	6118	1.3274	1.0	185.5688	10.3714
1.2173	15.0	6555	1.3282	1.0	180.7421	10.6344
1.1354	16.0	6992	1.3216	1.0	180.5052	10.6264
1.1011	17.0	7429	1.3290	1.0	181.7048	10.6168
1.0552	18.0	7866	1.3304	1.0	184.0366	10.7247
0.9884	19.0	8303	1.3401	1.0	183.3214	10.6767
0.965	20.0	8740	1.3503	1.0	181.0558	10.6815

Framework versions

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