Nvidia Rapids Dask CUDA Dataframes - 22x faster than Pandas! скачать в хорошем качестве

Nvidia Rapids Dask CUDA Dataframes - 22x faster than Pandas!

We use an 8GB GTX1070 to test and compare GPU and CPU Dataframes between: 🟪 NVIDIA Rapids Dask CUDA Dataframes (Dask cuDF) 🟪 NVIDIA Rapids CUDA Dataframes (cuDF), and 🟦 Pandas Dataframes. For this experiment, I created some CSV files of fake customer details from 1GB to 16GB, but intentionally put a lot of bad data in there like: 👎 Missing Values 👎 Wrong Values 👎 Incorrect Formatting I then wrote a Data Cleaning Python script in Jupyter Notebook/Jupyterlab that: 1️⃣ Loads the 1GB - 16GB CSVs into a Dataframe, 2️⃣ Performs 5 basic data cleaning steps, and 3️⃣ Exports the cleaned data back to CSV I then timed how long each library took to perform the Data Cleaning steps. The results for each dataset: 1:26 1GB 2:32 2GB 3:58 4GB 4:37 8GB & 16GB 4:53 Summary of Results 5:18 How To Install NVIDIA Rapids on Windows or Ubuntu Linux ---------- Nvidia Rapids is an accelerated data science framework for speeding up data science pipelines with an NVIDIA GPU (requires Pascal, Turing, Ampere, or better). It uses Python syntax and contains many awesome functions similar to Pandas Dataframes, and Sklearn (SciKit Learn). ---------- Subscribe to my YouTube channel: https://www.youtube.com/edgecate?sub_... ---------- #NvidiaRapids, #Dask, #Dataframes