Framework for training and tracking deep learning models built on pytorch-lightning and hydra. Soon to encompass both image classification, semantic segmentation and object detection.
Currently the plotting is a considerable amount of the config. Essentially we do coding with the config file. It would be better if the plotting had an interface in which we could create codified plotters.
The interface need to take in a container of tensors and need to be able to extract the correct tensors needed to perform the plotting.
Currently the plotting is a considerable amount of the config. Essentially we do coding with the config file. It would be better if the plotting had an interface in which we could create codified plotters.
The interface need to take in a container of tensors and need to be able to extract the correct tensors needed to perform the plotting.