We will create a new workspace and run the Amira notebook.
Open the new RAPIDS project, click Workspaces from the project menu.
Click + Create New Workspace.
For the Workspace Environment, choose RAPIDS from the dropdown.
For the Workspace IDE, choose JupyterLab. (Make sure you click on this or you won’t be able to launch)
For the Hardware Tier, choose GPU 6 cores - 45 GiB.
Click Launch Now.
Once the workspace starts, click Terminal from the launcher.
Run the following commands to copy the RAPIDS sample projects into the Domino filesystem:
mkdir /mnt/repo cp -r /rapids/notebooks/repos/cuml /mnt/repo
Close your Terminal window
Click File Changes from the Workspace menu
Click Sync All Changes.
Under Notebook in JupyterLab, click on Python 3 (ipykernel)
On the left directory structure, open repo -> cuml -> notebooks
Click on arima_demo.ipynb to open it
On the top, under Run, select Run all cells.
Scroll down through the notebook to explore the results.
There are a number of other notebooks within the folder. Check out this list to see if you any of them are interesting to you:
Forest Inference Demo: Save and load an XGBoost model into FIL and infer on new data.
KMeans Demo: Predict using k-means, visualize and compare the results with Scikit-learn’s k-means.
KMeans Multi-Node Multi-GPU Demo: Predict with MNMG k-means using dask distributed inputs.
Linear Regression Demo: Demonstrate the use of OLS Linear Regression for prediction.
Nearest Neighbors Demo: Predict using Nearest Neighbors algorithm.
Random Forest Demo: Use Random Forest for classification, and demonstrate how to pickle the cuML model.
Random Forest Multi-Node Multi-GPU Demo: Solve a classification problem using MNMG Random Forest.
Target Encoder Walkthrough: Understand how to use target encoding and why it is preferred over one-hot and label encoding with the help of criteo dataset for click-through rate modelling.
