Old Mistral v0.2 model to v0.3?

Have any one successfully convert Mistral 0.2 model to 0.3 with new v3 tokenizer and 32768 vocabulary size? There [https://github.com/mistralai/mistral-finetune](https://github.com/mistralai/mistral-finetune) script there expand vocabulary size of older models. This is my first time working with raw model layers. Just converting v0.2 to bin use script resize and convert back is not working for me, I'm clearly missing a a 1-2 steps. Will be grateful with suggestions and ideas.