hlibbabii / log-recommender-cli Goto Github PK

Real datasets are usually big and in our case are processed on the cluster. It would be beneficial to add a small dataset for testing purposes

Currently, to run a job with the runner script, the whole config file should be put into the param folder. Since often, the configs don't differ much, it makes sense to have a default config and for running a job, submit a 'patch', the difference between the default config and the desired configuration.

Currently, the names are composed of preprocessing params key-value pairs. However the resulting names are long and not readable enough, which can get even worse if more preprocessing params are added

nocom_1_nonewlinestabs_1_nostr_1_en_only_0_numspl_0_scspl_0_spl_0

Starting from nocom, all the column are shifted by one column to the right (see the picture).

This behaviour is not observed when the colors of 0's and 1's are changed to the default one.

Scripts include:

• classification dataset generation
  ...

TBD

hlibbabii/log-recommender#22