Closed alena-m closed 4 months ago
Hi, The default config files for the generation (as appears in run_generation_pipeline.py) should work well. The first failure that you get ('mix of unknown and multiclass targets'), is due to an issue when saving and loading the dataset (it converts the annotation ranks from strings to floats). I will upload in the coming days a fix for this bug. Meanwhile, if you are not loading dumps (load_dump should be an empty string) it should work well.
Hi, could you provide examples how should look like
config_ranking.yml
for generation tasks? With defaultconfig_ranking.yml
(as of now in main branch) it fails with:If I add
eval
section toconfig_ranking.yml
with function_name=ranking:then it fails with: