In my last post I discussed setting a roadmap for this semester in open source, and now with some ideas in place, it is time to move on. Moving on from here would be metaphorically choosing my car, meaning what I will be doing to travel down the roadmap, and achieve my goals. In this post I will discuss a few of the options I am currently looking at.
Staying On Plan
The three projects I researched last week were Tensorflow, Scikit-learn, and Keras. All of these projects were very viable options, but required some deeper digging to see which issues could be conquered. Of these three projects I discovered 3 issues; 1 from Tensorflow, and 2 from Keras.
After choosing the issues I wanted to work on, I knew the first step was asking/telling the community I would like to work on this, so I added a comment on the three issues I chose.
The first thing I realized after posting my comment was the flair/title I had been given after my previous contribution to Keras.
EarlyStopping should print mode details
In this issue, a user discussed their request to have more details be printed after EarlyStopping. This user used this feature in multiple places, and requested that “When the callback stops the training it doesn’t say anything about which metrics caused the stopping”.
This issue will require looking into the callbacks class of Keras, which contains the method for EarlyStopping. This feature will use an additional parameter as a flag to set if the method will print the details upon EarlyStopping
Calculating mean for flow_from_directory
This is a pre-processing issue for Keras, and has since been requested to be moved to a different repo, keras-preprocessing. This issue is used to incorporate additional functionality into the preprocessing, and used for standardization on a data set. This functionality exists but needs to be modified to be integrated with the DirectoryIterator.
GetTempFilename is not implemented
This issue is straightforward in nature, there is declaration for this method, but no implementation. The method name explains its feature and purpose as it should return the temporary filename being used. The user explains this to be an error that occurred: “error C4716: ‘tensorflow::io::GetTempFilename’: must return a value” and shows the logs in a screenshot.