slack splash

The Meeshkan Slack Integration

Back to the main documentation page

After installing Meeshkan on Slack and reading the python-client docs, you can start using the Meeshkan Slack integration out of the box. If you have not installed Meeshkan on Slack yet, click here. If you have not read the python client docs, click here. Otherwise, read on!


Installing the Slack Integration

If you haven't installed the Slack integration into your workspace yet, click here. The installation screen requires you to authorize the following permissions.

img

After completing the installation, sign into Slack and you will see, on the channel in which you installed the Slack bot, a message that resembles the following.

If you do not see this message or have any other difficulties during the installation process, please use the Intercom icon at the bottom of this page to get in touch with our help center.


Hello world

From a script

Run any script using the Meeshkan executor to see basic notifications about that script directly in your Slack Workspace. For example, by running the script below available in our examples repository:

$ meeshkan submit pytorch_mnist.py

You will see the following notification in your Slack workspace.

img

From a Jupyter notebook

Coming in January 2019

From a Jupyter notebook, you can signal the start of a training section using the following api.

import meeshkan
  # ... then, in the cell with your main loop
  meeshkan.register("my_notebook_identifier")

You will see the following notification in your Slack workspace.

img

Reporting

From a script

When submitting a job, use the report_interval flag to control the reporting interval. For example, to set a report_interval of every five minutes for a given job, use:

$ meeshkan submit --report_interval 300 pytorch_mnist.py

You will see the following notification in your Slack workspace.

img

From a Jupyter notebook

Coming in January 2019

From a Jupyter notebook, you can set the report_interval using the following API.

meeshkan.set_report_interval(300)

You will see the following notification in your Slack workspace.

img

Buttons

Sometimes, notifications will contain buttons that propose several different actions that can be taken in the training of a Machine Learning model.

img

In the example above, the user has the option to restart a job with a recommendation or to stop a job. Not pressing either button will always let a job running until its natural conclusion.

Coming in May 2019

Buttons can be pressed via groups of users in addition to the user who submitted the job, and other fine-grained controls for notification content can be controlled via a team dashboard or directly from Slack via slash commands.


The Meeshkan Bot

The Meeshkan Bot, addressable by calling @Meeshkan-iML in your Slack workspace, provides useful information and actionable recommendations regarding your Machine Learning jobs. In addition, the bot can respond to commands such as changing the learning rate of a job, redistributing compute power when possible or stopping a job.

img

In the example above, the user redistributes compute in a cluster running distributed several distributed training algorithms.

For enterprise clients, the Meeshkan Bot also acts as a direct-line to Meeshkan enterprise support for all account-related questions.

Coming in March 2019

The Meeshkan Bot is currently an alpha feature of Meeshkan, as its NLP capabilities are limited to ensure a fluid and error-free developer experience. If for any reason the bot cannot understand a request, it will be forwarded to a human agent at Meeshkan central and will be acted upon in a timely manner.


Slash commands

Coming in May 2019

A variety of slash commands will allow users to containerize and deploy machine learning models for direct testing on Slack.


Community + Questions

The Meeshkan Community workspace is a place where Meeshkan builders unite to work on projects such as Kaggle competitions. It is also where the beta version of our bot lives and early adopters can check out the newest bot features. To join the Meeshkan Community workspace, click on the link below. For all other questions, use the intercom widget on the bottom-right corner of this page to reach out to our support desk.

Join our Community