Environment variable for configuring the default downloading location

[sxjscience]

Currently all models/datasets will be downloaded to ~/.mxnet/... by default. We should add an environment variable to configure the constant. Sometimes the user may not have enough space under ~ and will need to download the dataset to another location.

[lixin4ever]

@sxjscience Have you solved this problem?

[leezu]

@lixin4ever you can usually pass the root='/some/path/ parameter to functions that download files (ie datasets/models). The change proposed by @sxjscience needs some (trivial) changes in the code, but no-one has done them yet. If you would like to contribute it, please let us know and we can provide guidance.

[lixin4ever]

@leezu I am happy to contribute to this project. How to start? Is there any guidance or documentation online?

[leezu]

@sxjscience @szha I suggest this change should also be applied to mxnet gluon, not only the gluon nlp package. In particular instead of choosing a environment variable such as GLUONNLP_HOME we could introduce MXNET_HOME.

@lixin4ever Great! We have some how to contribute documentation available.

For this issue I guess you could add a helper function _get_path(*args) to utils.py which would check if MXNET_HOME is defined and respectively join the looked up MXNET_HOME or default ~/.mxnet with the *args specified, returning the composed path. Then the functions that specify os.path.join('~', '.mxnet', ...) and their documentation need to be changed to make use of _get_path. But feel free to take a different route. What do you think?

[szha]

@leezu We can start using MXNET_HOME with this package first and propagate the change back to mxnet later on.

[sxjscience]

Thanks @lixin4ever for contributing. This has been solved by #106 .