| 8 Jun 2024 |
 hexa | there were two rocm maintainers some time ago | 22:34:09 |
| hexa | ideally they would chime in and clarify their commitment | 22:34:18 |
 connor (he/him) | In reply to @hexa:lossy.network
make sure your libraries build on python3.12, we're switching to 3.12 as the primary python version in this cycle Oof does PyTorch Dynamo even work with 3.12 yet? I haven’t checked since 2.1 I think | 23:05:58 |
| 9 Jun 2024 |
 SomeoneSerge (matrix works sometimes) | Hmm there's a PR starting with python3Packages.torch: but Ofborg didn't try building it, just shows 20 green (eval&c) checks | 00:36:09 |
| connor (he/him) | In reply to @glepage:matrix.org
connor (he/him) (UTC-5), in case you have a bit of available CPU time, could you please run a nixpkgs-review pr --post-result 317576 ?
(If you don't to that's fine ofc) Rerunning it by the way; got stuck for 20h+ on tensordict’s checkPhase :/ | 04:05:26 |
 Gaétan Lepage | In reply to @connorbaker:matrix.org
Rerunning it by the way; got stuck for 20h+ on tensordict’s checkPhase :/ Thanks. The tensordict problematic problematic test has been disabled in https://github.com/NixOS/nixpkgs/pull/318111 | 07:47:15 |
| Gaétan Lepage | I now get stuck on python312Packages.botorch | 07:47:27 |
|  shekhinah removed their display name she⯰khinah - traumaturgische Beratung. | 16:57:35 |
| connor (he/him) | I have a custom config I use specifically for nixpkgs-review you may like | 17:59:18 |
| connor (he/him) | https://gist.github.com/ConnorBaker/305b1aebd7ee74a258a616bbbd4dcd7b | 17:59:55 |
| shekhinah changed their profile picture. | 18:04:55 |
| Gaétan Lepage | In reply to @connorbaker:matrix.org
I have a custom config I use specifically for nixpkgs-review you may like Wow | 18:07:32 |
| Gaétan Lepage | So botorch did build for you ? | 18:07:41 |
| connor (he/him) | Yeah it did after I disabled checks for it | 18:15:10 |
| connor (he/him) | Just posted three variations of nixpkgs-review on the PR, https://github.com/NixOS/nixpkgs/pull/317576 | 18:15:45 |
| connor (he/him) | Looks good to me! | 18:15:49 |
| connor (he/him) | I'm going to try a run with nix-cuda-test real quick | 18:16:02 |
| Gaétan Lepage | Thank you so much ! | 18:16:45 |
| Gaétan Lepage | yes, for me it hangs in the tests... | 18:16:54 |
| Gaétan Lepage | If this is not the case on master we should probably investigate that ?
Anyway, considering that a vast majority of the downstream packages do still build fine, I would argue for merging this PR. | 18:17:49 |
| Gaétan Lepage | As a more general thought, I find very important to mark broken packages as such as it prevents us from diving in the nixpkgs-review failures every time to investigate whether a breakage is a regression or not. | 18:18:53 |
| connor (he/him) | Agreed; I can't do it fast enough which is why I've just got that config I use | 18:25:26 |
| connor (he/him) | If I succeed in running nix-cuda-test, are you okay with me merging it? | 18:30:56 |
| Gaétan Lepage | In reply to @connorbaker:matrix.org
If I succeed in running nix-cuda-test, are you okay with me merging it? Yes, I am OK with it. | 19:22:14 |
| Gaétan Lepage | I will go once again through the failures while this completes | 19:22:27 |
| hexa | In reply to @ss:someonex.net
Hmm there's a PR starting with python3Packages.torch: but Ofborg didn't try building it, just shows 20 green (eval&c) checks https://github.com/NixOS/ofborg/issues/577 | 19:30:37 |
| hexa | use python311Packages or python312Packages instead | 19:30:56 |
| Gaétan Lepage | SomeoneSerge (UTC+3) what is your opinion on merging the torch update as is ? | 19:47:12 |
| Gaétan Lepage | I am pretty confident in the absence of regression in this PR | 19:48:03 |
| connor (he/him) | Gaétan Lepage: have you had a chance to try training a model with torch.compile? | 20:27:04 |
