Closed MilesCranmer closed 9 months ago
Thanks. I think this makes sense. Alternatively, we could update the REBOUND API calls in spock so it works with REBOUND(x) version 4.x. There shouldn't be too many.
I can fix this by tomorrow
On Mon, Feb 12, 2024 at 6:36 AM Hanno Rein @.***> wrote:
Thanks. I think this makes sense. Alternatively, we could update the REBOUND API calls in spock so it works with REBOUND(x) version 4.x. There shouldn't be too many.
— Reply to this email directly, view it on GitHub https://github.com/dtamayo/spock/pull/24#issuecomment-1938797967, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA2ABFYV2O4TUQJ2R7WJA63YTISHHAVCNFSM6AAAAABDE2MMGGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMZYG44TOOJWG4 . You are receiving this because you were mentioned.Message ID: @.***>
Version 1.6.0 updates SPOCK to the REBOUND 4.0 API.
I've only done this for the code that is needed by the user to run SPOCK--if you wanted to retrain the model, there are probably some residual calls to old versions of REBOUND, which should be easy enough to fix.
I also resaved the XGBoost FeatureClassifier model (JSON file) using the latest version of XGBoost, since the old JSON format will be deprecated in XGBoost version 2.3, and checked that we get the same outputs from the resaved model.
Thanks for catching that!
@dtamayo I couldn't build the most recent version of spock so I used pypi-timemachine to build the working version from early last year. It seems it needs the
reboundx==3.7.2
andrebound==3.14.0
especially. I put other version fixes in place just in case.