This is exactly my problem, too, which I described above! If you find any solution, would be glad if you could share. 🙂 Of course, I also share mine when I get one.
Hi @<1724960464275771392:profile|DepravedBee82>
After
Starting Task Execution:
It will literally start the process running your code,
Can you send the full log of the Task? what is the code doing? which system is running the agent (i.e. Windows/Mac/Linux docker etc)
Thanks for the response @<1523701205467926528:profile|AgitatedDove14> ! The code is a small FMNIST test training job written in PyTorch Lightning. On my local job (laptop GPU, Windows) it completes in ~ 5min. On the server (Linux, H100s) it just hangs at Starting Task Execution: . Neither of these are in Docker.
I would expect to see the standard PL progress bars outputted to the console, but since nothing is outputted, so I'm not sure how to go about debugging this. I've attached the full logs for local and remote
Hi @<1724960464275771392:profile|DepravedBee82> , can you perhaps add a simple print at the start of your code before any import?
Ok so my train.py now looks like this:
print("Before import")
from pathlib import Path
import hydra
import lightning as L
import torch
from coolname import generate_slug
from omegaconf import DictConfig
from src.datasets import JobDataModule
from src.models import JobModel
from src.utils import LogSummaryCallback, get_num_steps, prepare_loggers_and_callbacks
from clearml import Task
for i in range(torch.cuda.device_count()):
print(torch.cuda.get_device_properties(i).name)
print("Before task")
task = Task.init(project_name="ClearML Testing", task_name="FMNIST")
task.set_repo(
repo="git@ssh.dev.azure.com:v3/mclarenracing/Application%20Engineering/ml-queue-test"
)
task.set_packages("requirements.txt")
print("After task")
And the log looks like this:
Starting Task Execution:
Before import
2024-07-19 09:06:09
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
Before task
So it looks like it's getting stuck at Task.init
If there was an SSL issue it should log to console right?
ClearML is hosted on an on-prem kube cluster and to get it to log locally I needed to append my company cert to the file located at certifi.where() . Do you think the same needs to be done for the Python installation for the worker?
If there was an SSL issue it should log to console right?
correct, also the agent is able to report, so I'm assuming configuration is correct
@<1724960464275771392:profile|DepravedBee82> could you try to put the clearml import + Task .init at the top of your code?
Hi @<1523701205467926528:profile|AgitatedDove14> , I reordered the imports:
from clearml import Task
print("Before task")
task = Task.init(project_name="ClearML Testing", task_name="FMNIST")
task.set_repo(
repo="git@ssh.dev.azure.com:v3/mclarenracing/Application%20Engineering/ml-queue-test"
)
task.set_packages("requirements.txt")
print("After task")
print("Before import")
from pathlib import Path
import hydra
import lightning as L
import torch
from coolname import generate_slug
from omegaconf import DictConfig
from src.datasets import JobDataModule
from src.models import JobModel
from src.utils import LogSummaryCallback, get_num_steps, prepare_loggers_and_callbacks
for i in range(torch.cuda.device_count()):
print(torch.cuda.get_device_properties(i).name)
And here's the output:
Environment setup completed successfully
Starting Task Execution:
Before task
Still looks like it's getting stuck at Task.init
This is so odd,
could you add prints right after the Task.init?
Also could you verify it still gets stuck with the latest RC
clearml==1.16.3rc2
Hi @<1523701205467926528:profile|AgitatedDove14> , here's my code with some more prints:
from clearml import Task
print("Before Task.init")
task = Task.init(project_name="ClearML Testing", task_name="FMNIST")
print("Before task.set_repo")
task.set_repo(
repo="git@ssh.dev.azure.com:v3/mclarenracing/Application%20Engineering/ml-queue-test"
)
print("Before task.set_packages")
task.set_packages("requirements.txt")
print("After task")
print("Before import")
from pathlib import Path
import hydra
import lightning as L
import torch
from coolname import generate_slug
from omegaconf import DictConfig
from src.datasets import JobDataModule
from src.models import JobModel
from src.utils import LogSummaryCallback, get_num_steps, prepare_loggers_and_callbacks
print("After import")
I've attached the full log (using RC2). Still getting stuck at Task.init - very weird
My understanding is that on remote execution Task.init is supposed to be a no-op right?
My understanding is that on remote execution Task.init is supposed to be a no-op right?
Not really a no-op, it would sync Argpasrer and the like, start background reporting services etc.
This is so odd! literally nothing printed
Can you tell me something about the node "mrl-plswh100:0" ?
is this like a sagemaker node? we have seen things similar where Python threads / subprocesses are not supported and instead of python crashing it just hangs there
It’s a Dell XE9680 rack server with 8xH100s which is located in our office, running AlmaOS. We have successfully run training jobs on it inside Docker (without ClearML) which work fine (will check with my team if we’ve got something to train without Docker). I’ve also tried different Python versions; 3.9 (Alma default) and 3.11 which you can see in the log above. It’s a really bizarre issue and outside of print statements I’m not really sure where to look.
You mentioned sync argparser & reporting, so I’ll try removing Hydra to rule that out, and other loggers in PL and see from there …
Hmm, I'm without, no reason why it will get stuck .
Removing all the auto loggers, this can be done with
Task.init(..., auto_connect_frameworks=False)
which would disconnect all the automatic loggers (Hydra etc) off course this is for debugging purposes
I've added that flag, removed all PL loggers & callbacks and all references to Hydra, but no luck 😞
Okay I have an idea, it could be a lock that another agent/user is holding on the cache folder or similar
Let me check something
Can you add before the Task.init
import os
print(os.environ)
I just ran with this in my local task, and all the env vars were printed to console, but in ClearML they are not in the console log. Presumably that's because it's printed before ClearML is logging?
Sorry, on the remote machine (i.e. enqueue it and let the agent run it), this will also log the print 🙂
Ah yes you were right, it does still print on remote. Here you go:
environ({'LANG': 'en_GB.UTF-8', 'PATH': '/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin', 'HOME': '/root', 'LOGNAME': 'root', 'USER': 'root', 'SHELL': '/bin/bash', 'INVOCATION_ID': '2cf51dc43b78470cb14c29f5f653ee18', 'JOURNAL_STREAM': '8:224108', 'SYSTEMD_EXEC_PID': '134947', 'PYTHONUNBUFFERED': '1', 'CUDA_DEVICE_ORDER': 'PCI_BUS_ID', 'CLEARML_WORKER_ID': 'mrl-plswh100:0', 'TRAINS_WORKER_ID': 'mrl-plswh100:0', 'CLEARML_CONFIG_FILE': '/tmp/.clearml_agent.vw6k62pq.cfg', 'TRAINS_CONFIG_FILE': '/tmp/.clearml_agent.vw6k62pq.cfg', 'CLEARML_TASK_ID': 'b0abe1da01bd4539a8e06699141c893a', 'TRAINS_TASK_ID': 'b0abe1da01bd4539a8e06699141c893a', 'CLEARML_LOG_LEVEL': 'INFO', 'TRAINS_LOG_LEVEL': 'INFO', 'CLEARML_LOG_TASK_TO_BACKEND': '0', 'TRAINS_LOG_TASK_TO_BACKEND': '0', 'PYTHONPATH': '/root/.clearml/venvs-builds/3.9/task_repository/ml-queue-test:/root/.clearml/venvs-builds/3.9/task_repository/ml-queue-test::/usr/lib64/python39.zip:/usr/lib64/python3.9:/usr/lib64/python3.9/lib-dynload:/root/.clearml/venvs-builds/3.9/lib64/python3.9/site-packages:/root/.clearml/venvs-builds/3.9/lib/python3.9/site-packages'})
@<1724960464275771392:profile|DepravedBee82> I just realized, the agent is Not running in docker mode, correct? (i.e. venv mode)
If this is the case how come it is running as root? (could it be is is running inside a container? how was that container spinned?)
Yes the agent is running in venv mode afaik. As for why it’s running as root - I’ll ask our engineer …
He confirmed that it’s not inside a container. Trying to figure out why it’s running as root but would it make a difference if it was? Is it better to run the agent from a user profile?
Edit: it might be a container! Just checking now...
Nope - confirmed to be running on the OS's Python environment, although he said that the agent was supposed to have it's own user - looking into that now
Nope - confirmed to be running on the OS's Python environment,
okay so bare metal root is definitely not recommended.
I'm not sure how/why it get's stuck though 😞
Any chance you can run the agent as non-root?
Also maybe preferred in docker mode, so it is easier for you to control the environment of the Task
Will try non-root and get back to you. I’m also trying to reproduce on a different machine too
@<1523701205467926528:profile|AgitatedDove14> we've now configured the server to have it's own user account to run the agent so it is no longer running as root, but no luck 😞
Before os.environ
environ({'LANG': 'en_GB.UTF-8', 'PATH': '/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin', 'HOME': '/home/clearml', 'LOGNAME': 'clearml', 'USER': 'clearml', 'SHELL': '/bin/bash', 'INVOCATION_ID': 'da8e36a03c7348efbb7db360755e92b3', 'JOURNAL_STREAM': '8:244189055', 'SYSTEMD_EXEC_PID': '1970812', 'PYTHONUNBUFFERED': '1', 'CUDA_DEVICE_ORDER': 'PCI_BUS_ID', 'CLEARML_WORKER_ID': 'mrl-plswh100:0', 'TRAINS_WORKER_ID': 'mrl-plswh100:0', 'CLEARML_CONFIG_FILE': '/tmp/.clearml_agent.4ll2u471.cfg', 'TRAINS_CONFIG_FILE': '/tmp/.clearml_agent.4ll2u471.cfg', 'CLEARML_TASK_ID': '4ab4c22b02ed4d1f86ff4fac663828f0', 'TRAINS_TASK_ID': '4ab4c22b02ed4d1f86ff4fac663828f0', 'CLEARML_LOG_LEVEL': 'INFO', 'TRAINS_LOG_LEVEL': 'INFO', 'CLEARML_LOG_TASK_TO_BACKEND': '0', 'TRAINS_LOG_TASK_TO_BACKEND': '0', 'PYTHONPATH': '/home/clearml/.clearml/venvs-builds/3.9/task_repository/ml-queue-test:/home/clearml/.clearml/venvs-builds/3.9/task_repository/ml-queue-test::/usr/lib64/python39.zip:/usr/lib64/python3.9:/usr/lib64/python3.9/lib-dynload:/home/clearml/.clearml/venvs-builds/3.9/lib64/python3.9/site-packages:/home/clearml/.clearml/venvs-builds/3.9/lib/python3.9/site-packages'})
Before Task.init
I managed to set up my (Windows) laptop as a worker and reproduce the issue. Would that suggest an issue with ClearML server?
Our server is deployed on a kube cluster. I'm not too clear on how Helm charts etc. work, but if there are any obvious things we should check, let me know and I can ask our DevOps engineer