Modin* Cheat Sheet

Recommended installation using conda (all backends)

conda install -c conda-forge modin-all

Recommended installation using conda (Ray* backend)

conda install -c conda-forge modin-ray

Recommended installation using conda (Dask* backend)

conda install -c conda-forge modin-dask

Installation using PyPI* (all backends)

pip install modin[all]

Installation using PyPI (Ray backend)

pip install modin[ray]

Installation using PyPI (Dask backend)

pip install modin[dask]

Installation using PyPI (MPI backend)

Switch to a Ray backend with a command prompt (if not enabled): for versions after 0.12

export MODIN_ENGINE=ray

Switch to a Ray Framework backend in the code (if not enabled): for versions after 0.12

import modin.config as cfg

cfg.StorageFormat.put(‘ray’)

import modin.pandas as pd

Switch to a Dask backend with a command prompt (if not enabled): for versions after 0.12

export MODIN_ENGINE=dask

Switch to a Dask backend in the code (if not enabled): for versions after 0.12

import modin.config as cfg

cfg.StorageFormat.put(‘dask’)

import modin.pandas as pd

Convert a Modin object to a Pandas object (example in bold)

import modin.pandas as pd     

df_log=pd.concat([self.df_log])

import pandas as pd

occ_dict = dict(df_log['EventTemplate']._to_pandas().value_counts())

df_event = pd.DataFrame()

        df_event['EventTemplate'] = df_log['EventTemplate'].unique()

Set number of cores Modin uses (Modin uses all available resources by default)

# set to Modin to only utilize 4 cores

export MODIN_CPUS=4