All pull requests

New pull request

Open

(18)

Closed

(0)

sbd: expose the Slurm GPU keys in create_blocks.py
#19dadamsncsa opened Sep 17, 2026Collaborator
slurm: split the wait limit into a queue limit and a run limit
#18dadamsncsa opened Sep 17, 2026Collaborator
[HELD — do not merge] sbd: seed the DE population with the CCSD ansatz, not a lone orbital rotation
#17dadamsncsa opened Sep 17, 2026Collaborator
sbd: cap concurrent solver jobs with a tag, not the worker pool
#16dadamsncsa opened Sep 17, 2026Collaborator
sbd: say when the object store was skipped, and let its name be set
#15dadamsncsa opened Sep 17, 2026Collaborator
sbd: let one failed walker fail one walker, not the whole run
#14dadamsncsa opened Sep 17, 2026Collaborator
slurm: run steps inside an allocation the driver already holds
#13dadamsncsa opened Sep 17, 2026Collaborator
slurm: submit a set of jobs as one array instead of one queue entry each
#12dadamsncsa opened Sep 17, 2026Collaborator
slurm: make the QPU directive spelling a site setting
#11dadamsncsa opened Sep 17, 2026Collaborator
slurm: fall back to sacct when scontrol cannot answer
#10dadamsncsa opened Sep 17, 2026Collaborator
slurm: stop the job when a module fails to load
#9dadamsncsa opened Sep 17, 2026Collaborator
slurm: grade a job's stderr by the job's outcome, not by the stream
#8dadamsncsa opened Sep 17, 2026Collaborator
slurm: do not leave a job queued when the wait is abandoned
#7dadamsncsa opened Sep 17, 2026Collaborator
slurm: let a block request GPUs
#6dadamsncsa opened Sep 17, 2026Collaborator
sbd: persist the raw device shots before they are filtered
#5dadamsncsa opened Sep 17, 2026Collaborator
sbd: add a simulator quantum source the DE loop can optimise
#4dadamsncsa opened Sep 17, 2026Collaborator
sbd: add a slurm target to create_blocks.py
#3dadamsncsa opened Sep 17, 2026Collaborator
sbd: import prefect_ray lazily so concurrent mode works without Ray
#2dadamsncsa opened Sep 17, 2026Collaborator

Navigation Menu

All pull requests

sbd: expose the Slurm GPU keys in create_blocks.py

slurm: split the wait limit into a queue limit and a run limit

[HELD — do not merge] sbd: seed the DE population with the CCSD ansatz, not a lone orbital rotation

sbd: cap concurrent solver jobs with a tag, not the worker pool

sbd: say when the object store was skipped, and let its name be set

sbd: let one failed walker fail one walker, not the whole run

slurm: run steps inside an allocation the driver already holds

slurm: submit a set of jobs as one array instead of one queue entry each

slurm: make the QPU directive spelling a site setting

slurm: fall back to sacct when scontrol cannot answer

slurm: stop the job when a module fails to load

slurm: grade a job's stderr by the job's outcome, not by the stream

slurm: do not leave a job queued when the wait is abandoned

slurm: let a block request GPUs

sbd: persist the raw device shots before they are filtered

sbd: add a `simulator` quantum source the DE loop can optimise

sbd: add a `slurm` target to create_blocks.py

sbd: import prefect_ray lazily so concurrent mode works without Ray

Pull requests