Closed hlieberman closed 1 month ago
Hi @hlieberman thank you very much for your contribution, I fixed it that for now smart-importer forces beancount < 3 and added the ci job on the PRs. Would you mind rebasing your PR so that the ci checks can run?
Not sure why Github decided that a force push should close this PR there, but... reopened and rebased. :)
There are many occasions when the training data set may include accounts which should not be predicted (e.g., accounts used for manual reconciliation of AR/AP). This feature allows the user to stop the predictor from learning these accounts, thus preventing contamination of the training set, without having to maintain a separate filtered copy of their transactions.
NB: Currently, all CI runs are broken because of beancount 3.0 hitting PyPi. Manual testing with beancount<3.0.0 shows all tests passing.