[[Kaggle]] offers both access to [[graphics processing unit|GPU]] and [[TPU]] in their notebook [[virtual machine]] environment for [[deep learning]] tasks. Turn on you preferred option under *Settings > Accelerator*. There is sometimes a waiting period of a few minutes to use the more performant units. You are limited to some hours of compute each week depending on your selection. If using TPU, you must configure the machine learning library to use the TPU and stream the data using a supported method. See [here](https://www.kaggle.com/docs/tpu) for details. Use `!nvidid-smi` to print the details of the GPU you are using. > [!Tip]- Additional Resources > - [Kaggle | Learn With Me: Getting Started with TPUs](https://youtu.be/1pdwRQ1DQfY?si=M8pWiztA9RLzzBZt)