When aeon is freshly built (pip install .) with all optional dependencies. It takes a lot of time to load.
Eg: Deep Clusterers have taken anywhere between 1 to 15 minutes in a Jupyter Notebook.
I would say the reason for this is a large dependency to other heavyweight packages like Tensorflow and also a large python code-base. All of these have to be loaded in the RAM for functions to be interpreted by the Python Interpreter.
Suggest a potential alternative/fix
Consider and Discuss adding support for lazy loading in the future (maybe as next year GSoC project). Lazy loading is a technique for waiting to load certain parts of a package until they are needed. Popular packages like the scikits use it.
TonyBagnall
commented
3 months ago
Describe the issue
When
aeonis freshly built (pip install .) with all optional dependencies. It takes a lot of time to load. Eg: Deep Clusterers have taken anywhere between 1 to 15 minutes in a Jupyter Notebook.I would say the reason for this is a large dependency to other heavyweight packages like Tensorflow and also a large python code-base. All of these have to be loaded in the RAM for functions to be interpreted by the Python Interpreter.
Suggest a potential alternative/fix
Consider and Discuss adding support for lazy loading in the future (maybe as next year GSoC project). Lazy loading is a technique for waiting to load certain parts of a package until they are needed. Popular packages like the
scikitsuse it.https://github.com/scientific-python/lazy-loader
Additional context
No response