Update README.md

README.md

@@ -34,4 +34,5 @@ The model works by:
 It was trained on a parallel corpus (corrupted + fixed sentence) with CTC loss. 
 On our test dataset, it reduces OCR errors by 41%.
 
+Training code: [here](https://github.com/slone-nlp/bashkort-spellcheker/blob/master/experiments/06_ctc_bert.ipynb). 
 Training details: in [this post](https://habr.com/ru/articles/744972/) (in Russian).