Use choice inputs

Install yq

Fix yml

Adjust yml structure

Add custom cascade

Fix defaults

Reduce inputs

Add more datasets

Add more datasets

Fix dataset name

Revert "Parametrize update benchmark too"

Parametrize update benchmark too

Use choice inputs

Use 1M laion-1m-no-filters for now

Use 1M cohere datasets

Use 2 small cohere datasets

Run test

Revert "Use matrix to clean machines"

Use matrix to clean machines

Fix parallel

Clean up machines in parallel

Clean up machines on failure

Fix

Simplify workflow

Sort by dataset

[pre-commit.ci] auto fixes from pre-commit.com hooks

Update workflow logic

Set 10 machines as default

Uncomment

[pre-commit.ci] auto fixes from pre-commit.com hooks