You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Data collection for training an additional spacekit ML model on "TAC" data (tso, ami, coron) is producing an unusually small number of L3 products. Despite collecting data on a daily basis from the pipeline over several months, the number of datasets needed to train the TAC model continues to be insufficient. Logs from the past several months show L1 inputs are being collected and identified correctly as "TAC", but very few of them are successfully matching to an L3 product, even accounting for time delays (e.g. if an L3 product takes a day or longer than its L1 inputs to complete).
Review existing TAC data (L1 and L3) collected so far
Review matching algorithm for TAC datasets
Determine if 2) the issue lies with algorithm or b) there really aren't L3 datasets to match with in the first place
If a: revise matching alg
If b: find out why this is the case and if it's expected
The text was updated successfully, but these errors were encountered:
Issue JP-3761 was created on JIRA by Ru Kein:
Data collection for training an additional spacekit ML model on "TAC" data (tso, ami, coron) is producing an unusually small number of L3 products. Despite collecting data on a daily basis from the pipeline over several months, the number of datasets needed to train the TAC model continues to be insufficient. Logs from the past several months show L1 inputs are being collected and identified correctly as "TAC", but very few of them are successfully matching to an L3 product, even accounting for time delays (e.g. if an L3 product takes a day or longer than its L1 inputs to complete).
The text was updated successfully, but these errors were encountered: