ClearML results page:
step 1
ClearML Monitor: GPU monitoring failed getting GPU reading, switching off GPU monitoring
Traceback (most recent call last):
File "/var/folders/1c/vz6m8k653j1_xpmpfgynshc00000gn/T/tmp4myzvumt.py", line 61, in <module>
task.upload_artifact(
File "/Users/almoghitelman/.pyenv/versions/sunbit-ai-312/lib/python3.12/site-packages/clearml/task.py", line 2341, in upload_artifact
raise exception_to_raise
File "/Users/almoghitelman/.pyenv/versions/sunbit-ai-312/lib/python3.12/site-packages/clearml/task.py", line 2322, in upload_artifact
if self._artifacts_manager.upload_artifact(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/almoghitelman/.pyenv/versions/sunbit-ai-312/lib/python3.12/site-packages/clearml/binding/artifacts.py", line 749, in upload_artifact
pickle.dump(artifact_object, f)
TypeError: cannot pickle '_thread.RLock' object
Launching the next 0 steps
Setting pipeline controller Task as failed (due to failed steps) !
pipeline completed
util class that write logs to txt file and holds the paths for data,model etc. @<1523701435869433856:profile|SmugDolphin23>
Hi @<1678212417663799296:profile|JitteryOwl13> ! Are you trying to return the logger from a step?
Hi @<1523701435869433856:profile|SmugDolphin23> , yes since I want to use it next steps
Thanks for the response, I tried to creat Logger.current_logger in each step but I got -
RuntimeError: can't create new thread at interpreter shutdown
Moving objects between steps is usually done via the artifacts mechanism. How are you building the pipeline, with decorators?
I’m getting error although logger_setup not calling ClearML logger, it’s just an inner class we use. cam we pass object between steps? (our objects) @<1523701070390366208:profile|CostlyOstrich36> @<1523701435869433856:profile|SmugDolphin23>
ValueError: Could not retrieve a local copy of artifact setup_logger
@PipelineDecorator.pipeline(name='Pipeline_Trail', project='Pipeline_Trail', version='0.1')
def main():
_args = params.parse_args()
setup_logger = init_experiment(_args)
data = load_dataset(_args, setup_logger)
train_model(_args, data, setup_logger)
Yes, passing custom object between steps should be possible. The only condition is for the objects to be pickleable. What are you returning exactly from init_experiment
?
the first step init Logger and return it object
Logger.current_logger()
Each step is a separate task, with its own separate logger. You will not be able to reuse the same logger. Instead, you should get the logger in the step you want to use it calling current_logger
Hi @<1678212417663799296:profile|JitteryOwl13> , can you please add a snippet that reproduces this?
How can we send objet from step1 to step2 without creating it again?
Your object is likely holding some file descriptor or something like that. The pipeline steps are all running in separate processes (they can even run on different machines while running remotely). You need to make sure that the objects that you are returning are thus pickleable and can be passed between these processes. You can try to see that the logger you are passing around is indeed pickalable by calling pickle.dump(s)
on it an then loading it in another run.
The best practice would be to have a separate logger for each step