Torch-snippets

🐞 `attach_hooks` will accept any custom hook

🎉 Make `markup2.AD.__repr__` the summary
🎉 Expose `markup2.AD` as `AD2`
🎉 Make `icecream` a requirement
🎉 Min python is 3.8

🎉 New IO hooks system in `torch_snippets.trainer.hooks`
🎉 Updated markup2.AD.summary and add print_summary methods

🎉 Experimental `AD` in torch_snippets.markup2 that infers variable names E.g. - `(p=10; AD(p) == {'p': 10})`
🐞 `isin` will not add +1 (useful for both absolute and relative boxes now)
🐞 `write_json` will support numpy, torch and AttrDict

🎉 add `find_address` to `AttrDict` that can return all path locations for a specific key
🎉 add `summary` to `AttrDict` that can give an outline of the dictionary
🎉 add `write_summary` to `AttrDict` that writes the summary to a textfile
🎉 `show` can now show bb colors `{"r": (255, 0, 0), "g": (0, 255, 0), "b": (0, 0, 255), "y": (255, 0, 255)}` if `df` has column called `color`
🎉 `AD` is an alias for `AttrDict`
🎉 `AD` can directly consume kwargs

🧹 import only important functins from `dates.py`
🎉 add `backup_all_notebooks` that backs up every notebook present in a specific folder
🎉 `reset_logger` can disable stdout logging if needed, using `disable_stdout=True` kwarg (False by default)
`common_items` will take a list of folders and return common stems from the folders
images will show a black border when grid is True