Closed astrojuanlu closed 3 years ago
That's fair, I'd support an upgrade in testing to test if the speed/memory usage is still as high now. You're running 4.6.1 IIRC and Conda's latest release is 4.10.x. We could fallback to Mamba otherwise until the new alternative scheme is implemented and shipped. My reasoning to try Conda again is basically to make sure the Conda community is getting what they are expecting and reduce the risk of having to deal with subtle implementation differences between Mamba and Conda.
As to using Mamba for all of RTD, I should note that Anaconda and a few other community members like @wolfv and me are currently ramping up work on Conda on making Conda's solver pluggable to use Mamba solver instead. That is one of many new initiatives to breath life into the Conda Open Source project again and better cater to Conda users, commercial or community-based.
You're running 4.6.1 IIRC and Conda's latest release is 4.10.x.
For some reason, the latest release of miniconda is 4.9.2: https://docs.conda.io/en/latest/miniconda.html#linux-installers
I also checked https://repo.anaconda.com/miniconda/ and I found that there is a 4.10.x version but it's not available for x86_64
. So, I'm upgrading to 4.9.2 in #174 for now. @jezdez is that enough?
We upgraded Miniconda3 to 4.9.2 in #174, closing!
Originally posted by @jezdez in https://github.com/numba/numba/issues/7182#issuecomment-874631554
It is unlikely that we will rebuild our
stable
andlatest
images, because that would break lots of builds. We are working on an alternative scheme but we haven't started with the implementation yet. In the meantime, we could update conda in ourtesting
image to something more recent, but at the same time, we are internally debating whether to do a more widespread mamba rollout, since it is faster and consumes less RAM.Thoughts @readthedocs/core ?