Closed h-mayorquin closed 2 years ago
In general I would think all we really need is (a) version of nwb-conversion-tools
, and (b) version of spikeextractors
. Those are the API's that have the most changes related to debugs, function calls, etc.
While the full frozen state would indeed be required to perfectly (100%) ensure reproducibility of original files, this might be overkill for our purposes.
In general I would think all we really need is (a) version of
nwb-conversion-tools
, and (b) version ofspikeextractors
. Those are the API's that have the most changes related to debugs, function calls, etc.While the full frozen state would indeed be required to perfectly (100%) ensure reproducibility of original files, this might be overkill for our purposes.
@CodyCBakerPhD
The changes that we used in both nwb-conversion-tools
and spikeextractors
are not included in any tagged version of them, aren't they? (i.e. the latest additions to cellexplorersortingextractor
in spikeextractors).
@h-mayorquin No, probably not; you can pin it to the commit hash of the version state used successfully to run the conversion though (example: https://github.com/catalystneuro/nwb-conversion-tools/blob/main/requirements-full.txt#L13-L14)
@h-mayorquin No, probably not; you can pin it to the commit hash of the version state used successfully to run the conversion though (example: https://github.com/catalystneuro/nwb-conversion-tools/blob/main/requirements-full.txt#L13-L14)
@CodyCBakerPhD
The latest commit follows your suggestion above to only have the pinned versions of nwb-conversion-tools
and spikeextractors
.
@h-mayorquin Looks great!
This is to discuss how (or if) we should store the
requirements.txt
(for pip) and theenvironment.yml
(for conda). I added the frozen version of the state of the repos just after I ran the conversion.Right now, they just stand as an indicator of what libraries the environment that I used in dandi had. I am not sure this will be sufficient to reproduce the environment as I have not tested it but maybe this should be enough?
Let me know what you think.