| 14 Oct 2024 |
 GaƩtan Lepage | In reply to @ss:someonex.net
Seems like dependencies failed to build: https://hercules-ci.com/accounts/github/SomeoneSerge/derivations/%2Fnix%2Fstore%2Fn3lww4jsfan66wyryh3ip3ryarn874q5-onnxruntime-1.18.1.drv?via-job=e51bf1d4-6191-4763-8780-dd317be0b70b
Rather than debugging this, I'd advise you look into https://hydra.nix-community.org/job/nixpkgs/cuda/onnxruntime.x86_64-linux. There hasn't been any official announcements from nix-community's infra team to the best of my knowledge -> no "promises", but the hope is that this will become the supported and long-term maintained solution
This is building the cuda version of onnx ? | 19:03:19 |
 SomeoneSerge (back on matrix) | Yes but also the hydra history is all green š¤· | 19:08:54 |
| GaƩtan Lepage | Yes, weird... | 19:13:19 |
| SomeoneSerge (back on matrix) | Noticed https://github.com/SomeoneSerge/nixpkgs-cuda-ci/issues/31#issuecomment-2412043822 only now, published a response | 19:22:08 |
| GaƩtan Lepage | I can't get onnx to build...
Here are the logs in case someone know what is happening: https://paste.glepage.com/upload/eel-falcon-sloth | 20:08:13 |
| SomeoneSerge (back on matrix) |
error: downloading 'https://github.com/abseil/abseil-cpp/archive/refs/tags/20230125.3.tar.gz' failed
lol
| 20:19:08 |
| SomeoneSerge (back on matrix) | In reply to @ss:someonex.net
Yes but also the hydra history is all green š¤· Maybe that just came in from staging | 20:19:30 |
| 15 Oct 2024 |
 connor (he/him) | In reply to @glepage:matrix.org
I can't get onnx to build...
Here are the logs in case someone know what is happening: https://paste.glepage.com/upload/eel-falcon-sloth Onnx's CMake isn't detecting at least one dependency, so it tries to download them all in order, starting with abseil. Since there's no networking in the sandbox, it fails. | 00:06:48 |
| connor (he/him) | I'm currently working on Onnx packaging for a thing, and you can see what I've got going on here: https://github.com/ConnorBaker/cuda-packages/blob/main/cudaPackages-common/onnx.nix
(It's a combination C++/Python install so it's gnarly. But better than having two separate derivations with libraries built with different flags, I guess.) | 00:09:04 |
| GaƩtan Lepage | Ok interesting, thanks for sharing | 05:46:57 |
| GaƩtan Lepage | Is your plan to upstream this to nixpkgs ? | 05:47:13 |
| GaƩtan Lepage | [triton update]
triton-llvm fails during the test phase.
Logs: https://paste.glepage.com/upload/fish-jaguar-pig | 08:48:05 |
|  atagen joined the room. | 11:38:21 |
| SomeoneSerge (back on matrix) | In reply to @glepage:matrix.org
[triton update]
triton-llvm fails during the test phase.
Logs: https://paste.glepage.com/upload/fish-jaguar-pig Can't reproduce, builds for me | 12:35:31 |
| SomeoneSerge (back on matrix) | In reply to @glepage:matrix.org
[triton update]
triton-llvm fails during the test phase.
Logs: https://paste.glepage.com/upload/fish-jaguar-pig * Can't reproduce, builds for me. Maybe we tried different HEADs? | 12:36:26 |
| atagen | hi, what am I missing to get a cache hit? going by this hydra output torch should be in the cache (for nixpkgs 5633bcf). I have nix-community cachix set up, allowUnfree, cudaSupport,, and the package in question is providing its overlay properly with final.callPackage so it ought to be using my system packages | 12:46:24 |
| atagen | https://gist.github.com/atagen/615e187e323f3ca3f5f9d40e55ce2b7c | 12:55:50 |
| atagen | oof, could it be because I'm specifying python311Packages instead of python3Packages? | 12:57:30 |
| atagen | ... yup, that was it | 12:58:23 |
| SomeoneSerge (back on matrix) | https://github.com/NixOS/nixpkgs/blob/70f9c111b27db0d459a227e477acce62016cbf10/pkgs/top-level/release-cuda.nix#L118 | 13:04:59 |
| SomeoneSerge (back on matrix) | In reply to @glepage:matrix.org
[triton update]
triton-llvm fails during the test phase.
Logs: https://paste.glepage.com/upload/fish-jaguar-pig With the current HEAD and ccache off I just reached the pytest branch | 14:17:17 |
| GaƩtan Lepage | In reply to @ss:someonex.net
With the current HEAD and ccache off I just reached the pytest branch You mean that you were able to build it fine ? | 14:47:18 |
| SomeoneSerge (back on matrix) | Yes | 14:47:27 |
| SomeoneSerge (back on matrix) | Well the pytest bit fails with these 20 tests ofc but that'll come later | 14:47:41 |
| GaƩtan Lepage | Ok, weird then... | 14:49:24 |
| GaƩtan Lepage | Btw, I'm running a cross-system review for this triton PR. | 14:49:35 |
| GaƩtan Lepage | quite a few rebuilds | 14:49:40 |
| connor (he/him) | In reply to @glepage:matrix.org
Ok interesting, thanks for sharing Yep, that's the goal. My hope is to replace the current CUDA packaging stuff with what I've got there.
I personally will be maintaining CUDA 11.8 for a while but mark it as end of life. Since it requires toolchains which will be removed upstream, I'll keep it out of tree.
My plan is to only maintain the latest version of CUDA, but block upgrades to newer versions if some prominent packages don't build, even on master.
I plan to ship the same version of most libraries that NVIDIA does with its ML containers, which means roughly a monthly release cadence. | 16:19:57 |
| connor (he/him) | Of course, all this is pending agreement with the other maintainers, but it would certainly help cut down the scope of CUDA packages and allow us to better populate the cache since there'd be really just one version supported upstream | 16:20:36 |
| GaƩtan Lepage | This looks smart indeed ! | 16:55:13 |
