PyCharm 2018.3.7 (Professional Edition) on macOS 10.14.6
I started a new project with a remote interpreter.
When I try to debug remotely I encountered errors that I can't understand the source.
File "/home/nlp/user_name/.pycharm_helpers/pydev/_pydevd_bundle/pydevd_comm.py", line 1501, in do_it
var_obj = pydevd_vars.getVariable(self.thread_id, self.frame_id, scope, attrs)
File "/home/nlp/user_name/.pycharm_helpers/pydev/_pydevd_bundle/pydevd_vars.py", line 227, in getVariable
var = resolver.resolve(var, k)
File "/home/nlp/user_name/.pycharm_helpers/pydev/_pydevd_bundle/pydevd_resolver.py", line 203, in resolve
return getattr(dict, key)
And it goes on many more errors like so.
In the same Pycharm exists another older project which connects to the same remote machine that has no problem debugging remotely. I tried to start a new project within the Pycharm and I also created another remote python virtual env, both didn't solve the problem, I still see the same errors.
In the same trouble making project when I switch the interpret to a local one everything works fine.
Can you help with that, please?
Please sign in to leave a comment.
Another important comment on the problem:
It has to do with variables store in CUDA, when I run the process remotely only on CPU then things work properly but when the remote process uses CUDA it gets this problem.
Hi, I think the best course of actions is to submit an issue to https://youtrack.jetbrains.com/issues and provide:
1. Minimal code/project sample to reproduce the issue
2. Logs from **Help | Collect Logs and Diagnostic Data**
3. Full output from the debugger console
This way we can analyze it more efficiently.
Thanks,
Solved the problem.
The error was due to data-parallel handling.
When you try to debug your code on a GPU card it is better to use only a single GPU card.