Closed bakhtos closed 1 month ago
Please bring the boilerplate back. As you said it breaks compatibility with existing scripts. The point of the boilerplate was also to keep the models away from the scripts. We can definitely use the new generic train method as well as it gives more control to user.
Also the conda should work when using https://github.com/EvoTestOps/LogLead/blob/main/environment_no_dl.yml that needs no DL libraries or HW. Has worked for multiple people so far.
Please bring the boilerplate back. As you said it breaks compatibility with existing scripts. The point of the boilerplate was also to keep the models away from the scripts. We can definitely use the new generic train method as well as it gives more control to user.
If you mean the changes proposed in #19 , we can discuss them later after this PR is finished. In this PR I do not change the internal implementations at all, just move stuff around.
Rename AnomalyDeection to AnomalyDetector
Rename LCSMap to SpellParser
Change vec_name to be the vectorizer object instead of the name to lookup
Remove BertEmbeddings from parsers
Solves #5 Enhancers, parsers, and loaders can now be imported as
AnomalyDetection class can be imported as
Same with other classes:
I have fixed the imports in all test and demo scripts, but I have not run them because I had issues setting up the conda environment. So someone should test all existing scripts + some additional ones you might have just to be sure.
I suggest making a release
1.0.0
from the current state ofmain
branch, merge this pull request and then immediately make a2.0.0
release since these refactorings break compatibility with existing scripts.