There are then three checks until a pull request is merged:
automatic (see below)
Chris: to accept the run
Sachi: after the data has been collected and the assembly is available in Brain-Score
Automatic checks:
no naming conflicts between registered StimulusSets
all registered functions actually create a StimulusSet in the desired format (i.e. pandas DataFrame with at least id column, as well as pointing to an image file that can be loaded through e.g. PIL)
notes:
could ignore StimulusSets that are registered and already available on S3
Further points to discuss:
are experimental requests separate from the StimulusSet? are they part of it? should users register both a stimulus set as well as an experimental request?
My favorite CI providers are:
There are then three checks until a pull request is merged:
Automatic checks:
id
column, as well as pointing to an image file that can be loaded through e.g. PIL) notes:Further points to discuss: