Repository files navigation

[WIP] Provides helper commands related to an ML research Slack group. Mostly related to a slurm cluster.

Available commands:

• /cluster - Get summary of the nodes in the cluster

Create a Slack bot app with appropriate scope (chat:write,command,im:history,users:read) and get the Slack tokens. Subscribe to eventsapp_home_opened,message.im. You can find more information on how to do thishere.

SLACK_BOT_TOKEN andSLACK_SIGNING_SECRET environment variables are required to run the app.

• ForSLACK_BOT_TOKEN copy theBot User OAuth Access Token under theOAuth & Permissions sidebar.
• SLACK_SIGNING_SECRET is available in your app'sBasic Information page underApp Credentials.

export SLACK_BOT_TOKEN=xoxb-your-tokenexport SLACK_SIGNING_SECRET=your-signing-secret

conda env create -f environment.yaml

Follow the instructionshere

export SLURM_INCLUDE_DIR=/usr/includeexport SLURM_LIB_DIR=/usr/lib64git clone https://github.com/PySlurm/pyslurm.git && cd pyslurmgit checkout v23.2.0pip install .

Run it on a node with access to the cluster.

python app.py

• The cluster GUI is shameless rip-off ofslurm_web. If you are looking for a web GUI for cluster profiling, check it out.
• slurm_gpustat