!eWOErHSaiddIbsUNsJ:nixos.org

NixOS CUDA

276 Members
CUDA packages maintenance and support in nixpkgs | https://github.com/orgs/NixOS/projects/27/ | https://nixos.org/manual/nixpkgs/unstable/#cuda57 Servers

Load older messages


SenderMessageTime
5 Nov 2025
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)Added to merge queue02:07:23
@apyh:matrix.orgapyh
In reply to @connorbaker:matrix.org
Why are you building for so many CUDA capabilities? I can’t really think of a reason you’d need that range in particular.
's a distributed ml training application that needs to run on everything from gtx 10xx gpus to modern data center GH/GB200s :/
03:27:37
@apyh:matrix.orgapyhmost common hardware is gonna be 30xx 40xx 50xx, h100, a100, b20003:27:56
@apyh:matrix.orgapyhthough.. i could just see what pytorch precompiled wherls runs on and limit to that 03:28:54
@apyh:matrix.orgapyhshould be fine03:28:56
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)Flash attention doesn’t support anything older than Ampere I thought03:29:07
@sporeray:matrix.orgRobbie Buxton V2 does 03:29:19
@sporeray:matrix.orgRobbie Buxton V3 is hopper only 03:29:24
@apyh:matrix.orgapyhya its only v3 iirc03:29:26
@sporeray:matrix.orgRobbie Buxton V4 (cute) is Blackwell 03:29:33
@sporeray:matrix.orgRobbie BuxtonBut that’s a wip03:29:38
@apyh:matrix.orgapyhand yeah fair enough I could drop fa for older gpus, ig i can provide cuda capabilities different per package03:30:06
@sporeray:matrix.orgRobbie Buxton Requires cutlass-dsl 03:30:07
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)V2 doesn’t support older than Ampere per their readme, unless they forgot to update it03:30:25
@apyh:matrix.orgapyh
In reply to @connorbaker:matrix.org
V2 doesn’t support older than Ampere per their readme, unless they forgot to update it
yeah makes sense then maybe I'll just drop and see if anyone complains lol
03:30:43
@sporeray:matrix.orgRobbie Buxton
In reply to @connorbaker:matrix.org
V2 doesn’t support older than Ampere per their readme, unless they forgot to update it
It’s not optimized but it runs
03:30:46
@sporeray:matrix.orgRobbie Buxton V3 is busted 03:30:54
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)So long as you have PTX it should be fine03:30:55
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8) cries in Ada 03:31:14
@connorbaker:matrix.orgconnor (burnt/out) (UTC-8)Robbie have you been following PyTorch Helion? Seems interesting (but boooooo Bazel build)03:32:07
@sporeray:matrix.orgRobbie Buxton Yeah seems cool, morach too! 03:32:37
@sporeray:matrix.orgRobbie Buxton * Yeah seems cool, monarch too! 03:32:47
@apyh:matrix.orgapyhmonarch looks sweeet03:32:52
@apyh:matrix.orgapyhwere any of you at pytorch conference?03:33:02
@apyh:matrix.orgapyhshoulda asked when i was there lol03:33:08
@sporeray:matrix.orgRobbie BuxtonNah03:33:33
@sporeray:matrix.orgRobbie BuxtonWas it any good!?03:33:41
@apyh:matrix.orgapyhnah03:33:44
@apyh:matrix.orgapyhlol03:33:45
@apyh:matrix.orgapyhcouple neat talks but nothing revolutionary 03:33:57

Show newer messages


Back to Room ListRoom Version: 9