A bunch of notebooks and scripts experimenting with Caffe, TF, Keras, PyTorch etc.
- Added PyCrayon, Now you can create a client an experiment and pass it to the Experiment class - https://github.com/torrvision/crayon
- Re-wrote the experiment class and separated model, dataset from it and left the control to user
- Added a notebook walking through how it can be used.
- My new found Love - PyTorch!
- Included script to train/finetune any of the Imagenet model in PyTorch Model Zoo with any Image Dataset
Integrate with pycrayon and check logging on TensorBoardFix the pretrain - freeze - unfreeze - classes dependency hellCreate a Tutorial Notebook and include CitationsWrite Documentation for the arguments for the script- Maybe we can create factory methods for the optimizer, scheduler, criterion so that
the script can become declarative and configurable with just arguments.