I am working through the with some friends at the Seattle Go Center.
I have noticed with some of the users more experienced with Go than with programming, the current working directory required to get the code to work is often confusing.
The following change makes each program find its filesystem data relative to its own program path. That way, programs will always find their data files so long as the code and data is kept together, as it is in the git repository.
Note that one data file, mnist.pkl.gz, is in a subdirectory of code/dlgo. That means the code can be put into a pip package and dissociated from the data. So it appears mnist.pkl.gz is already broken in the pip package. I'm going to semi-arbitrarily choose to include mnist.pkl.gz in this commit.
See #12 for underlying questions on the defacto requirements for mnist.pkl.gz.
jeffhgs
commented
5 years ago
I am working through the with some friends at the Seattle Go Center.
I have noticed with some of the users more experienced with Go than with programming, the current working directory required to get the code to work is often confusing.
The following change makes each program find its filesystem data relative to its own program path. That way, programs will always find their data files so long as the code and data is kept together, as it is in the git repository.
Note that one data file, mnist.pkl.gz, is in a subdirectory of code/dlgo. That means the code can be put into a pip package and dissociated from the data. So it appears mnist.pkl.gz is already broken in the pip package. I'm going to semi-arbitrarily choose to include mnist.pkl.gz in this commit.
See #12 for underlying questions on the defacto requirements for mnist.pkl.gz.