1 Aug 2024 |
SomeoneSerge (utc+3) | The reference used not to be retained in the outputs | 17:20:02 |
SomeoneSerge (utc+3) | Trying 24.05 | 17:21:01 |
SomeoneSerge (utc+3) | I think it's time to add an exportReferencesGraph test to e.g. torch, or better yet to a few core packages | 17:22:47 |
SomeoneSerge (utc+3) | As a very unambiguous way to ensure that this stuff isn't referenced | 17:23:10 |
SomeoneSerge (utc+3) | Oh wait. Actually, now it is going to be in the closure if we include triton | 17:23:31 |
SomeoneSerge (utc+3) | I think we keep a reference to the toolchain in triton | 17:24:03 |
SomeoneSerge (utc+3) | A a rough estimate, I think 23.11 is a good commit xD | 17:29:22 |
SomeoneSerge (utc+3) | Sorry got to leave now for a while | 17:29:35 |
yorickvp | bisecting the following:
let
pkgs = import ./. {
config = {
allowUnfree = true;
cudaCapabilities = [ "8.6" ];
cudaSupport = true;
};
};
in
{
torchtest = (pkgs.python3.pkgs.torch.override { openai-triton = null; }).overridePythonAttrs (o: {
disallowedReferences = [ pkgs.python3.pkgs.torch.cudaPackages.cuda_nvcc.stdenv.cc.cc.lib ];
USE_CUDNN = 0;
USE_KINETO = 0;
USE_QNNPACK = 0;
USE_PYTORCH_QNNPACK = 0;
USE_XNNPACK = 0;
INTERN_DISABLE_ONNX = 1;
ONNX_ML = 0;
USE_ITT = 0;
USE_FLASH_ATTENTION = 0;
USE_MEM_EFF_ATTENTION = 0;
USE_FBGEMM = 0;
USE_MKLDNN = 0;
});
}
| 17:37:11 |
yorickvp | In reply to @yorickvp:matrix.org
bisecting the following:
let
pkgs = import ./. {
config = {
allowUnfree = true;
cudaCapabilities = [ "8.6" ];
cudaSupport = true;
};
};
in
{
torchtest = (pkgs.python3.pkgs.torch.override { openai-triton = null; }).overridePythonAttrs (o: {
disallowedReferences = [ pkgs.python3.pkgs.torch.cudaPackages.cuda_nvcc.stdenv.cc.cc.lib ];
USE_CUDNN = 0;
USE_KINETO = 0;
USE_QNNPACK = 0;
USE_PYTORCH_QNNPACK = 0;
USE_XNNPACK = 0;
INTERN_DISABLE_ONNX = 1;
ONNX_ML = 0;
USE_ITT = 0;
USE_FLASH_ATTENTION = 0;
USE_MEM_EFF_ATTENTION = 0;
USE_FBGEMM = 0;
USE_MKLDNN = 0;
});
}
disallowedReferences seems not to work, though | 17:52:16 |
yorickvp | it was already broken on dc7b3febf8d862328d8704de5c8437d2df442c76 | 18:02:02 |
yorickvp | (23.11 branchoff) | 18:02:06 |
SomeoneSerge (utc+3) | Ehh cuda_nvcc.stdenv.cc looks wrong | 18:35:31 |
SomeoneSerge (utc+3) | Or maybe you're tight it's the unwrapped one | 18:36:12 |
SomeoneSerge (utc+3) | * Or maybe you're right it's the unwrapped one | 18:36:19 |
SomeoneSerge (utc+3) | Butthe reference could be through triton | 18:36:48 |
SomeoneSerge (utc+3) | I think you have to print-rpath | 18:37:09 |
yorickvp | 23.05 uses a cuda gcc that matches the default gcc, so I can't tell if it works there | 18:38:51 |
yorickvp | okay, the reason this worked before for me is that I was using a torch binary that was also compiled with gcc 12 lib | 20:54:03 |
yorickvp | I think this reproduces my problem: https://gist.github.com/yorickvP/28713f755f720e7f23d7f64a02a45ca3 after configurePhase, build.ninja's LINK_PATH entries all have gcc-12.3.0-lib | 21:12:28 |
2 Aug 2024 |
hexa (UTC+1) | Download image.png | 20:35:44 |
hexa (UTC+1) | killing botorch tests on x86-64-linux when | 20:35:56 |
hexa (UTC+1) | they're just flaky when run under pressure | 20:49:54 |
hexa (UTC+1) | https://github.com/NixOS/nixpkgs/pull/331857 | 21:23:41 |
SomeoneSerge (utc+3) | In reply to @hexa:lossy.network killing botorch tests on x86-64-linux when 💯 | 23:16:15 |
SomeoneSerge (utc+3) | In reply to @hexa:lossy.network killing botorch tests on x86-64-linux when Negative affinity in Nix when | 23:17:27 |
5 Aug 2024 |
hexa (UTC+1) | I think disabling that particular test helped | 12:19:49 |
connor (he/him) (UTC-7) | botorch test suite 🙅♂️ | 16:44:17 |
aidalgol | I misread that as "brotorch" for a sec. | 19:20:03 |
aidalgol | * I misread that as "brotorch" for a sec. | 19:20:12 |