pytorch_fundamentals

Learning objectives:

• Learn how to use Tensors with CPUs and GPUs
• Understand how to manage, scale and normalize your datasets
• Build an image recognition model using a neural network
• Learn how to optimize a model
• Learn how to enhance model inference performance

Setup

Install SkyPilot - makes the process of provisioning a machine easy:

pip install -U "skypilot[aws]"

Configure AWS access:

pip install boto3
aws configure

To get a machine with a GPU, I run the following command:

sky gpunode -p 8888 -c jupyter-vm --gpus T4:1 --cloud aws --region us-west-1 --use-spot

Then, on the VM, I activate the pre-configured by AWS environment:

source activate pytorch

After, I start the Jupyter notebook server:

jupyter notebook

Finally, I select kernel in my local (Codespaces) Jupyter notebook to match the http address with the token generated by the command above.