This is the base PyTORCH project.
Contains:
- Jupyter for interactive development
- Examples of training tasks
Repository pytorch-examples contains example of spatial transformation using MNIST dataset. If you want to see the details about transformation, please look at jupyter notebook spatial_transformer_tutorial.ipynb in the src directory using Jupyter after project installation.
Steps to run spatial transformer as a separate task or on GPU:
- Install project with default source repo and config
- Go to tab "Tasks"
- Choose "standalone" -> "worker"
- (Optional) Adjust resource requests to use GPU: set GPU to 1
- Click "Save and execute"
- Then, watch the logs in tab "Logs": expand row named "standalone:X"
- Wait unitl job is succeeded
- Open "Jupyter" tab and look into training dir: there should be 2 .png files containing a set of test images and a set of corresponding processed images. You can download them and see the result of transformation.