metadata

license: mit
language:
  - ru
  - kbd
widget:
  - text: Я иду домой.
    example_title: Я иду домой.
  - text: Дети играют во дворе.
    example_title: Дети играют во дворе.
  - text: Сколько тебе лет?
    example_title: Сколько тебе лет?
  - text: На следующий день мы отправились в путь.
    example_title: На следующий день мы отправились в путь.

m2m100_ru_kbd_44K

This model is a fine-tuned version of facebook/m2m100_418M on a ru-kbd dataset, containing 44K sentences from books, textbooks, dictionaries etc.. It achieves the following results on the evaluation set:

Loss: 0.9399
Bleu: 22.389
Gen Len: 16.562

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
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
2.2391	0.18	1000	1.9921	7.4066	16.377
1.8436	0.36	2000	1.6756	9.3443	18.428
1.63	0.53	3000	1.5361	10.9057	17.134
1.5205	0.71	4000	1.3994	12.6061	17.471
1.4471	0.89	5000	1.3107	14.4452	16.985
1.1915	1.07	6000	1.2462	15.1903	16.544
1.1165	1.25	7000	1.1917	16.3859	17.044
1.0654	1.43	8000	1.1351	17.617	16.481
1.0464	1.6	9000	1.0939	18.649	16.517
1.0376	1.78	10000	1.0603	18.2567	17.152
1.0027	1.96	11000	1.0184	20.6011	16.875
0.7741	2.14	12000	1.0159	20.4801	16.488
0.7566	2.32	13000	0.9899	21.6967	16.681
0.7346	2.49	14000	0.9738	21.8249	16.679
0.7397	2.67	15000	0.9555	21.569	16.608
0.6919	2.85	16000	0.9441	22.4658	16.493

Framework versions

Transformers 4.21.0
Pytorch 1.10.0+cu113
Datasets 2.4.0
Tokenizers 0.12.1