- Complete the PyTorch Blitz tutorial.
- Create a neural network to classify the preloaded data in the Jupyter Notebook provided
- Create data loader
- Create model
- Create criterion
- Data loop over epochs and batches
- Iterate over the minibatches in random order
- Reset gradient and perform forward pass, backpropogation, and optimizer step
- Validate on test set
- Push your code to submit your work
- Attend a mentor session to defend your implementation and answer questions
- If you do not want to use a Jupyter Notebook, use these commands to import the data:
data_train = torchvision.datasets.MNIST('./', download=True, train=True, transform = transform)
data_test = torchvision.datasets.MNIST('./', download=True, train=False, transform = transform)
- Please pay attention when reading through the tutorial because the mentors will be asking you questions
- If you have issues with installation, post on Piazza first so that others can view possible solutions