if I put ~/clearml in the default_output_uri key, and start the task, when run as agent in GCP I get clearml.Task - INFO - Completed model upload to file:///$github_proj_directory/~/clearml/$proj_name/$experiment_name

no, only in the clearml.conf file
Now I removed the output_uri in the conf file of the machine that started the task, and when I run it as agent in GCP it works.
Is this a bug?

My setting is the following:
-run script in tl2 (local server)
-clone task and enque it, run it in GCP
Ideally I would like to:
if the script in run in tl2 it should save to local filesystem if the script is run in GCP it should save to GS

I'm not sure I understand - can you share the different settings you used in you clearml.con file (just put the relevant parts of the file here, if you can) and explain where should each setting point to?

Do you have anything in the script itself related to output_uri ?

if I write:
in tl2 conf : default_ output_uri : "/home/tglema/clearml"
in GCP it saves in that same dir

even though I have set the default_output_uri in GCPs conf file

How exactly did you add it to the cleaml.conf file?

is that clearer?

let me run the model_upload example in your repo instead of my script

since it didn't change (the packages installed)

this is the one in tl2

so with these two configurations, and no output_uri in the task creation in the script:
I get model saved model in tl2 and in GCP (when run as agent):
/home/tglema/git_repo/~/clearml/

sure 🙂

do I create an issue for this as well? SuccessfulKoala55 ?

and just reuse it

Ok, here it is 🙂 https://github.com/allegroai/clearml/issues/298

Can I see the exact lines you added to the .conf file?

I'm not sure...

in GCP it saves in that same dir
even though I have set the default_output_uri in GCPs conf file

You mean that in GCP it stores in "/home/tglema/clearml" even though you've set
default_ output_uri : "gs://..." in the clearml.conf file on the GCP instance?

mmm..I'm having the same issue:
in my GCP agent:

base) root@gst-cv-glema3-final-tf2-cu101:~# clearml-agent daemon --queue redness
Current configuration (clearml_agent v0.17.1, location: /root/clearml.conf):

sdk.storage.cache.default_base_dir = ~/.clearml/cache
sdk.development.default_output_uri =

ok, ran the script, and had the same issue..I detected another bug I think..going to put it outside the thread

now I have:
in tl2: default_ output_uri : "~/clearml"
in GCP: default_ output_uri : "gs://..."

Can I see the whole config file? (you can hide any sensitive info there, of course)

if I launch the same script in GCP, (I don't run it as a clearml-agent), then everything works fine

Please do 🙂

this is the one in GCP

See here: https://github.com/allegroai/clearml-agent/blob/822984301889327ae1a703ffdc56470ad006a951/docs/clearml.conf#L288

but it still saves using the output_uri from the server that created the task

and in the script I removed the output_uri= , in the task initialization

is there a way to prevent from creating a new setup in my worker each time?