
Reputation
Badges 1
46 × Eureka!Yeah I had the same issue: https://clearml.slack.com/archives/CTK20V944/p1664887550256279
Here are the versions: WebApp: 1.7.0-232 • Server: 1.7.0-232 • API: 2.21
I can make an issue in GitHub if that helps keep track of it
Thanks for the response, I don't have any specific reason. I just wanted to have a something cleaner. We don't have much projects yet, so we just get these examples in the way. But it's not bad, I was just wondering. I'll remember to check for the environment variables for our next ClearML install. Thanks anyways, I won't take the trouble of removing them then
I had the same issue, they fixed it in 1.7.2rc1 I think. You can try pip install clearml==1.7.2rc1
Nope sorry I don't have that, but this was our discussion about it: https://clearml.slack.com/archives/CTK20V944/p1665477602050259
Yeah me too, I had no idea what was happening 😅
From what I've seen, another fix would be to pass a default value to your parameters, for example: def executing_pipeline(start_date=0, end_date=0):
Apparently, the issue is that it will always pass None
if there's no default value
No sorry, I found the where the logs are. And it doesn't seem to have any errors in the logs:
` [2022-10-14 17:22:50,771] [9] [INFO] [clearml.service_repo] Returned 200 for tasks.get_all in 3ms
[2022-10-14 17:22:50,784] [9] [INFO] [clearml.service_repo] Returned 200 for tasks.get_by_id in 7ms
[2022-10-14 17:22:50,853] [9] [INFO] [clearml.service_repo] Returned 200 for events.add_batch in 182ms
[2022-10-14 17:22:50,874] [9] [INFO] [clearml.service_repo] Returned 200 for tasks.edit in 28ms
[202...
Sooo for the SHA2 generation, I've tested 2 very different CPUs, and it makes a HUGE difference 😅 I probably have to upgrade my server
But this works strangely:
` @PipelineDecorator.component(cache=False, execution_queue="default")
def get_param():
return 'hello'
@PipelineDecorator.component(cache=False, execution_queue="default")
def get_best_model(task_ids):
import ...
print('task_ids:', task_ids, type(task_ids)) # task_ids: None <class 'NoneType'>
...
@PipelineDecorator.pipeline(
name='...',
project='...',
version='0.1'
)
def pipeline_entry(task_ids: List[str], ...):
print(task_ids, typ...
Okay, thanks
My computer just did an automatic suspend, as simple as that
Just dropping this here but I've had some funky compressions with very small datasets! It's not a big issue though, since it's still small and doesn't really affect anything
Do you know where I can find the logs for that?
With default settings, to upload 2 datasets of 120 GB and 70 Gb it took more than 6 hours! And this is to upload the dataset on the server itself, the upload pipeline is done on the same computer as clearml
Thanks for trying to help me out! Here's some code that should reproduce the error (at least, it did for me): https://github.com/allegroai/clearml-agent/issues/111
We've updated everything now, launched a new experiment and we're tracking the logs. I'll tell you if I find anything
From what I could see, generating SHA2:
i7-10700K: ~ 10 - 15 minutes Xeon E3-1240: 4 - 5 hours!Then in both cases I still have about an 1h30 to upload the images to the fileserver. Which I also find quite a bit slow, but the ClearML fileserver is on my old Xeon. I plan to upgrade my server and to test it again
If you have any ideas as to what could go wrong, I'd be happy to look at it. But since my venv is rebuilt at each new agent run, I'm really struggling to debug it
Weeell it seems to work with version 1.7.0 and not with 1.7.1
CostlyOstrich36 This looks like a bug? Here's a simpler version of it and what I'm getting:
` from clearml.automation.controller import PipelineDecorator
@PipelineDecorator.component(cache=False)
def step_one(my_arg):
print('step_one/my_arg:', my_arg) # step_one/my_arg: None
# I should not get None here! At least that's what I'm expecting
@PipelineDecorator.pipeline(name='custom pipeline logic', project='examples', version='0.0.5')
def executing_pipeline(my_arg):
print('my_ar...
Yes sure CostlyOstrich36 , I'm just trying to pass some arguments from my __main__
to my pipeline_entry()
to my component get_best_model()
. But for some reason, I'm getting None
into get_best_model
instead of what I've given it in pipeline_entry
So this seems like it could work as a work-around:
` Python 3.10.6 (main, Aug 10 2022, 11:40:04) [GCC 11.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
import numpy as np
a = np.ones((100, 100, 3))
a.take(range(40), 0).take(range(40), 1).shape
(40, 40, 3)replaces a[0:40, 0:40] `