Try and make all reads able to read from URLs instead of just files.

GateNLP / gateplugin-LearningFramework

A plugin for the GATE language technology framework for training and using machine learning models. Currently supports Mallet (MaxEnt, NaiveBayes, CRF and others), LibSVM, Scikit-Learn, Weka, and DNNs through Pytorch and Keras.

GNU Lesser General Public License v2.1

26 stars 6 forks source link

Turns out this is not possible or difficult for many instances:

obviously for writing we need proper files
for wrappers we need an actual file directory even for application in order to export the instances and also because the wrapper cannot access the model from a non-file. This could get worked around by storing both temporarily in some scratch directory instead.
for LibSVM the API method requires a file when loading a model. The only workaround there would be to replace the API method by our own

GateNLP / gateplugin-LearningFramework

Try and make all reads able to read from URLs instead of just files. #52