Remote
SMT
@bioinfhotep@genomic.social
mastodon
4.6.6
Sequences and consequences.
Credit Pic : Cellular landscape cross-section through a eukaryotic cell, by Evan Ingersoll
0
Followers
0
Following
13
Posts
Joined November 06, 2022
Posts
Replying to
@eddelbuettel@mastodon.social
@eddelbuettel direct hit by this since i use a r2u containers that installed a version of this.path with this issue. I though i could continue living at the edge but with recent changes, i should version these containers even on staging !
#Rstats gurus how one check before crashing on CRAN check esoteric (or not so esoteric) issues around mac-old-release setup and other macos flavors not available on r-lib actions?
Last check of a package I submitted (works fine on macbuilder and other macos flavours)
https://www.r-project.org/nosvn/R.check/r-oldrel-macos-arm64/Rtinycc-00install.html
Replying to
@sebffischer@mastodon.social
@sebffischer@mastodon.social will check it out today and let you know !
Replying to
@sebffischer@mastodon.social
@sebffischer@mastodon.social this was the latest one indeed, 7 GB compressed size
Replying to
@sebffischer@mastodon.social
@sebffischer@mastodon.social yeah, this is my understanding too. maybe the docs should make this clearer though
Replying to
@bioinfhotep@genomic.social
@sebffischer@mastodon.social same class of issues for the container too, maybe other nvidia images to make the drivers available. This is usually a pain, i remember similar shenanigans with some GPU accelerated bioinformatics software where you needed to containers on the target machine.
But i think you are right, i did not install cuda12.8 from the mlverse repo
Replying to
@sebffischer@mastodon.social
@sebffischer@mastodon.social i first installed the dependencies including cuda deps, i managed to go through and solve a few of these by installing additional cuda stuff manually, but i stopped after this one. Lot of stuff in libpjrt_cuda.so from goml are dynamically linked it appears. I guess if i had read carefully the actions and dockerfiles i could do it more reliably. I downloaded anvil-cuda:latest, i will see if that works
Replying to
@sebffischer@mastodon.social
@sebffischer@mastodon.social after installing lot of nvidia stuff on wsl and seing stuff like (so and so is missing from libpgrt), i decided to use the image :D
I think the way to go about it on Colab is to reproduce the steps of the image on the instance and hope there is good caching. Also this is a way to test on TPUs
Replying to
@sebffischer@mastodon.social
@sebffischer@mastodon.social don t know about bayesian methods in general, but if i have time and since i have some Codex Sub now, i will try to do some ragebait induced NeoedgeR and compare to the numba using python port of edgeR https://brianrepko.github.io/blog/posts/2026-02-26-edgepython-and-ai/ https://github.com/pachterlab/edgePython
Testing on gpu should be possible since i have a gaming PC now (and i manage to install huge NVIDIA binaries on WSL)
But it would be great to have scripts for setting up anvil dev environment for testing on say google collab
Replying to
@jonny@neuromatch.social
@jonny@neuromatch.social ah, tour great perpetual draft specification. Noticeable is that it fits in 60 pages !
Replying to
@dgkf@fosstodon.org
@dgkf@fosstodon.org on static analysis, browsing Duncan Temple Lang's github (mostly for C FFI stuff), found reference to this package https://github.com/nick-ulle/rstatic they recommend in the meantime this https://github.com/flowr-analysis/flowr maybe the workers behind flowr might appreciate/need/want some type of funding/wider recognition