Mi!
Mi!
Looking at the traceback the problem is a with `libcrypto` and not kite.
Update on the prediction times for a day of data. ``` Old: Average time 1.5563322 This: Average time 0.9315222499999999 ```
@yetinam, ready for review.
I think the wheel is being built rn. No changes are required to the pipeline. Maybe we can tackle cibuildwheel in another PR?
@ThomasLecocq, with millions of traces, we eliminate quite some overhead. In SeisBench, we see a ~20% performance increase when annotating 3000 channels of five-minute length. ``` SeisBench annotating 3000 channels...
I have changed `UTCDateTime`'s signature to use keyword arguments instead of opaque `**kwargs`. The initialisation of `UTCDateTime` is very hot: at least 4x per trace creation and also during comparisons...
@ThomasLecocq, what is missing to merge this PR? Thanks for your review!
`typing.MutableMapping` is deprecated since 3.9. Consider using `collections.abc.MutableMapping` https://docs.python.org/3/library/typing.html#typing.MutableMapping