READMEpypi.md

Telegram bot ML stats.

This repository contains code for simple telegram bot for logging ML metrics.

Package description.

This package contains two classes to log ML metrics.

TGTableSummaryWriter

This writer is usefull for logging hyperparameters optimization or when you need log only numeric values.
Example:

For more information see TGTableSummaryWriter example.

TGImgSummaryWriter

This writer could be used for logging neural networks metrics per epoch.
Example:

For more information see TGImgSummaryWriter example.

Setup ml bot.

1. Create telegram bot. Save token.
2. Create public telegram group.
3. Add your bot to this group. Make him an administrator.
4. Run following code

   from tg_bot_ml.utils import create_credentials
   
   create_credentials(BOT_TOKEN, PUBLIC_CHAT_NAME, './credentials.yaml')
   

5. Make your group private.

In the end you will have file credentials.yaml that you can use to write logs into this group.

For example :

from tg_bot_ml.table_bot import TGTableSummaryWriter

if __name__ == '__main__':
    tg_bot = TGTableSummaryWriter('../credentials.yaml', 'Catboost Hyperopt : Customer Churn')
    tg_bot.add_record(lr=1e-3, n_estimators=100, MSE=1.3)
    tg_bot.send(sort_by="MSE", ascending=False)