Print problem when importing this project

[UserAEH]

My problem is that no print is displayed in the command line when using this module:
A simple example could show this problem:

from super_gradients.training import models
print("hello")

but no print appears on the console.
I tried with sys.stdout.write("hello") and sys.stdout.flush() but no results.
Is there any conflict with default print function? Do you have any idea about how to solve this issue?

[dagshub]

Join the discussion on DagsHub!

[Louis-Dupont]

Hi @UserAEH , can you try:

stdout = sys.stdout

from super_gradients.training import models

sys.stdout = stdout

print("hello")

[UserAEH]

Hi @UserAEH , can you try:

stdout = sys.stdout

from super_gradients.training import models

sys.stdout = stdout

print("hello")

It works perfectly, thanks.
Do you think there's some problem within the package or my environment? (In case there's a problem with the package, we could request a fix to enhance the project)

[Louis-Dupont]

It's the first time someone faced this issue, I guess there must be a conflict between SuperGradients and something else in your environment. Did you import anything before the snippet you showed, or it happened when just importing super_gradients?

[UserAEH]

No, there's nothing before the snippet. I posted the full code.
I'm working with Python 3.10.11 if you want to test it.

[Louis-Dupont]

Tried on 3.10.11 as well but it doesn't reproduce on my device.
I will investigate when I have some time, in the meantime I guess the workaround is to wrap the first super-gradient import with it. I think that the following imports won't create this issue.

You can also try the following code, it might work as well:

import sys
from super_gradients.training import models

sys.stdout = sys.__stdout__

print("hello")

[UserAEH]

Tried on 3.10.11 as well but it doesn't reproduce on my device. I will investigate when I have some time, in the meantime I guess the workaround is to wrap the first super-gradient import with it. I think that the following imports won't create this issue.

You can also try the following code, it might work as well:

import sys
from super_gradients.training import models

sys.stdout = sys.__stdout__

print("hello")

This solution is working. It seems like the problem is coming from the file common/abstractions/mute_processes.py.
When i commented the line 30:
#sys.stdout = open(os.devnull, "w")

Print worked with the first code i posted and without reassigning original value to stdout.
Another thing to mention is that there was one line that was added in the console before printing hello which i think is normal:

The console stream is logged into some\path\logs\console.log.

[Louis-Dupont]

Hi @UserAEH , I think this PR should fix it #1044 can try to install from master:
pip install git+https://github.com/Deci-AI/data-gradients

[UserAEH]

Hi @Louis-Dupont , i tried to install from master. The encoding is added but the issue persisted.

[ukoehler]

Problem confirmed for
super-gradients 3.1.1
Platform Windows 10

from super_gradients.training import Trainer is enough to cause the problem for me.

Workaround sys.stdout = sys.stdout works, but it is still annoying.

[pcismyname]

same problem here (Window 10)!
I discovered that my output shows up in the console where I open Jupyter lab.

[xcweric]

I found the solution:
(super_gradients 3.1.3)
\common\abstractions\abstract_logger.py
commented the line 9&18: 'mute_subprocesses()'
Then log and print() will show again:)
by default this func is always called which not friendly for some users like me:( , I spent a lot of time to find it...

[UserAEH]

I found the solution: (super_gradients 3.1.3) \common\abstractions\abstract_logger.py commented the line 9&18: 'mute_subprocesses()' Then log and print() will show again:) by default this func is always called which not friendly for some users like me:( , I spent a lot of time to find it...

The problem in your solution is that It prints also logger messages so I am not sure if it is a normal behavior. Logger should be written in log file not printed in the CLI.

[cobaltautomationdev]

this issue still exists until now 2024.01.14