Hello! I have been running into a curious error using pydra version 0.22 installed into a conda environment on the UW HPC ("Hyak").
A minimal example that demonstrates this is the following script:
import os.path as op
from tempfile import mkdtemp
import pydra
scratch_dir = "/gscratch/escience/arokem/"
scratch_dir_tmp = op.join(scratch_dir, "tmp_")
cache_dir_tmp = mkdtemp(prefix=scratch_dir_tmp)
@pydra.mark.task
def task(subject):
print("I am doing something really simple")
subject_list = ["01", "02"]
t = task(subject=subject_list, cache_dir=cache_dir_tmp).split("subject")
with pydra.Submitter(plugin="slurm",
sbatch_args=f"-J task -p gpu-a40 -A escience --mem=58G --time=10:00:00 -o /gscratch/escience/arokem/logs/task.out -e /gscratch/escience/arokem/logs/task.err --mail-user=arokem@uw.edu --mail-type=ALL") as sub:
sub(runnable=t)
When run it raises the following error:
Traceback (most recent call last):
File "simple.py", line 23, in <module>
sub(runnable=t)
File "/gscratch/escience/arokem/afq/lib/python3.8/site-packages/pydra/engine/submitter.py", line 42, in __call__
self.loop.run_until_complete(self.submit_from_call(runnable, rerun))
File "/gscratch/escience/arokem/afq/lib/python3.8/asyncio/base_events.py", line 616, in run_until_complete
return future.result()
File "/gscratch/escience/arokem/afq/lib/python3.8/site-packages/pydra/engine/submitter.py", line 80, in submit_from_call
await self.expand_runnable(runnable, wait=True, rerun=rerun)
File "/gscratch/escience/arokem/afq/lib/python3.8/site-packages/pydra/engine/submitter.py", line 128, in expand_runnable
await asyncio.gather(*futures)
File "/gscratch/escience/arokem/afq/lib/python3.8/site-packages/pydra/engine/workers.py", line 299, in _submit_job
self.error[jobid] = error_file.replace("%j", jobid)
AttributeError: 'NoneType' object has no attribute 'replace'
However, interestingly, the jobs are submitted and are executed with no further issues, so this is not actually causing any "real" problems for us (except for looking like an error might have happened).
Hello! I have been running into a curious error using pydra version 0.22 installed into a conda environment on the UW HPC ("Hyak").
A minimal example that demonstrates this is the following script:
When run it raises the following error:
However, interestingly, the jobs are submitted and are executed with no further issues, so this is not actually causing any "real" problems for us (except for looking like an error might have happened).