[New Config] Follow-up & Known issues

[HAOCHENYE]

Limitations

The unsupported syntax for New Config, and these syntaxes will not be supported in the future version

1. The configuration file name must comply with the naming convention of Python modules, which can only contain letters, numbers, and underscores, and cannot start with a number.
2. When importing variables from the base configuration file, such as from ._base_.alpha import beta, the alpha here must be the module (module) name, i.e., a Python file, rather than the package (package) name containing __init__.py.

Developing

Sytaxes are not supported now but could be supported in the future.

1. Functions and classes cannot be defined in the configuration file.

2. Importing multiple variables simultaneously in an absolute import statement, such as import torch, numpy, os, is not supported. Multiple import statements need to be used instead, such as import torch; import numpy; import os.

3. Cannot put imported module in a list directly like this

   from torch.nn import Linear, Conv2d
   a = dict(args=[Linear, Conv2d])

4. Dump the class object of Pure Python style config to pure text style, and recover the string back to the class object.
   Pure python style config

   from torch.nn import LInear
   model = dict(type=Linear)

   Dumped cfg

   model = dict(type='torch.nn.Linear')

   Load the dumped cfg

   print(Conig.fromfile(dumped_cfg).model)   # is a string, not nn.Linear

[wangg12]

Currently, (recursively) instantiating/building a class or a function without registries is not supported.

It could be supported though, as the document claimed and similar instantiation has been supported in d2's config.