ML and GPU Playgrounds
Train models on your own CSV, and learn how GPU kernels run, without installing anything. Runs in your browser, and your files never leave your device.
scikit-learn PlaygroundUpload a CSV and compare logistic regression, a decision tree and gradient boosting. No install and no code.
TensorFlow PlaygroundUpload a CSV, train a real neural network in your browser and watch the loss, the accuracy and the decision boundary.
CUDA PlaygroundWrite a CUDA kernel and watch every thread run: threadIdx, blockIdx, bounds errors, races and warp divergence. Works without a GPU.
About these playgrounds
The scikit-learn and TensorFlow playgrounds train on a CSV file you choose. The scikit-learn page compares classic models with a few clicks, and the TensorFlow page trains a small neural network and shows how its decision boundary changes. Both read the file in your browser.
The CUDA playground is a simulator. It runs a kernel on a small grid of threads in your browser, so you can see indexes, out-of-bounds writes, races and warp divergence without a GPU.