cuDF feels like Pandas to many but just faster owing to its GPU compatibility.
Installing cuDF on linux is way easier than on windows.
Here is a way to install cuDF on google colab:
The data used in the exploration is described here.
If you are new to G-colab, don’t forget to save the notebook to your drive and choose GPU runtime as follows:
You can use nvidia-smi command to know the allocated GPU hardware as shown in the notebook.
The methods of cuDF are Pandas like. Refer the cheatsheet from RAPIDS to know more. Furthermore, during any stage of Exploratory Data Analysis (EDA), you can convert cuDF dataframe to pandas dataframe i.e. from cudf.core.dataframe.DataFrame to pandas.DataFrame and back. Simple functions like .to_pandas() and .from_pandas() would do the trick.
Writers: Piyush Kulkarni, Apoorv Kumar, Rahul Kumar Raman
