Closed h1alexbel closed 2 months ago
@volodya-lombrozo take a look, please
@volodya-lombrozo yes, definitely. I didn't check the speed of code in @BeforeAll
section that loads a model. 16-19ms for tests execution, and ~3seconds for model loading. I will implement some sort of caching to speed up it a bit. Is 1.5 second of model load will be considered as acceptable? WDYT?
@h1alexbel Yes, it would be great. 1.5
seconds is acceptable. Please, go ahead.
@volodya-lombrozo updated pull request, take a look again, please
First run should take ~3 seconds, model will be cached in src/test/resources/ml/cached.bin
. Next test executions should take no longer than 1.5 seconds
@volodya-lombrozo updated pull request. Take a look again, please
@volodya-lombrozo updated pull request. Take a look again, please
@rultor merge
In this pull I've implemented
CachedModelSource.java
to cache model download inModelSourceInternet.java
. Enabled all unit tests related to ml.closes #101
PR-Codex overview
The focus of this PR is to update exception handling and introduce caching for model sources in the ML module.
Detailed summary
ModelSource.java
CachedModelSource.java
andCachedModelFs.java