Closed sebffischer closed 7 months ago
In mlr3torch, there are somewhat complicated "tensor-columns" that reference environments. Because the DataBackend hashes the whole data.table, this can lead to problems because of byte-code compilation (https://github.com/mlr-org/mlr3/blob/78ad4dd439bdc7ebfa2ca327eb6cb6e08c424e16/R/DataBackendDataTable.R#L159C24-L159C54).
mlr3torch
By merging this PR, it is possible to customize which information from a column is used to calculate the hash.
sebffischer
commented
7 months ago sebffischer
commented
7 months ago
In
mlr3torch, there are somewhat complicated "tensor-columns" that reference environments. Because the DataBackend hashes the whole data.table, this can lead to problems because of byte-code compilation (https://github.com/mlr-org/mlr3/blob/78ad4dd439bdc7ebfa2ca327eb6cb6e08c424e16/R/DataBackendDataTable.R#L159C24-L159C54).By merging this PR, it is possible to customize which information from a column is used to calculate the hash.