jmakov

> In this case compare_models(fit_kwargs={'class_weight': dict(zip(df['species'].unique(), [1,2,3]))} the result is [] Also my experience with 3.3.2.

@Midnighter perhaps this example will help you: https://stackoverflow.com/a/73871567

Um... how is this not a priority bug? Am I using (micro)mamba wrong?

Same experience here, doesn't seem to work: ```bash $ export CODON_PYTHON=/home/toaster/mambaforge-pypy3/envs/puma-lab/bin/python # this results in the same error # $ export CODON_PYTHON=/home/toaster/mambaforge-pypy3/lib/libpypy3.9-c.so $ export PYTHONHOME=/home/toaster/mambaforge-pypy3/envs/puma-lab $ condon run bin/script.py Python...

@arshajii thanks for the quick response! `puma-lab` is just the name of my `mamba Python3.9` environment (no idea why `codon` sees Python3.10 (though it is installed on the system, along...

@thobai as a workaround you can use filesystem based compression e.g. BTRFS with ZSTD-9.

Perhaps `.set_at_idx()` partially solves the issue since it accepts lists for indices and values. Also how is updating a column in a data frame in 10 lines of code in...

Any updates on this? Still doesn't work using `infer_schema_length=0` or `=None`.

I think what we're trying to do is called "streamed json parsing". Did somebody perhaps do some benchmarks how doing partial parsing with `gjson` compares to dedicated streaming json parsers...

Perhaps one could just wrap the models around https://github.com/scikit-learn-contrib/MAPIE ?