We are trying to build flash-attn reliably for more than one CUDA architecture. We recently were approved to use the cirun-openstack-cpu-xlarge runner and have merged the corresponding PR.
The maintainers of the feedstock @carterbox and @weiji14 are listed in the access json. I have double checked that the id numbers are correctly listed in the conda-forge-users.json.
However, the post-merge main branch and other open PRs fail to build. There is an error message from Cirun that users are not authorized.
{
"error": "User not authorized for the requested runners, user not present in 'users' list."
}
The github actions for "Build conda package" shows jobs waiting in the queue. Is the queue full? Is there somewhere that I can view the queue? Is the user access list out of sync? Thanks!
@Quansight/open-gpu-server
We are trying to build flash-attn reliably for more than one CUDA architecture. We recently were approved to use the cirun-openstack-cpu-xlarge runner and have merged the corresponding PR.
The maintainers of the feedstock @carterbox and @weiji14 are listed in the access json. I have double checked that the id numbers are correctly listed in the conda-forge-users.json.
However, the post-merge main branch and other open PRs fail to build. There is an error message from Cirun that users are not authorized.
The github actions for "Build conda package" shows jobs waiting in the queue. Is the queue full? Is there somewhere that I can view the queue? Is the user access list out of sync? Thanks!