Can't work with gpu(cuda)

[mfussenegger]

Can't work with gpu(cuda)

[sisrfeng]

I have given up and use pudb for python.

[sisrfeng]

@sisrfeng Does pudb work well with PyTorch? (probably the last question, thanks!)

Mostly.
But it sometimes may fail when using multiple GPUs or multi-process.

[mfussenegger]

No stopped threads. Cannot move

Indicates that the debugger is not stopped at a breakpoint, but still running. Maybe you just have to wait longer, or it's waiting for some input? I have no idea what input.cuda() does.

If it really isn't progressing, it's probably an issue with debugpy itself, not with nvim-dap-python

[nyngwang]

@sisrfeng Did you solve it eventually? I'm also setting up nvim-dap for PyTorch recently. It will be great if you could share some insights on this. Thanks!

[nyngwang]

@sisrfeng Does pudb work well with PyTorch? (probably the last question, thanks!)

[nyngwang]

@sisrfeng Could you share your config for PyTorch? This is the command I use without considering debugging:

python main.py --epoch 200 -b 128 --lr 0.05 --gpu 0 --gc

Should I create a new configuration for my case?

[sisrfeng]

@sisrfeng Could you share your config for PyTorch? This is the command I use without considering debugging:

python main.py --epoch 200 -b 128 --lr 0.05 --gpu 0 --gc

Should I create a new configuration for my case?

Without considering debugging, I think a configuration is not needed. What error do you see?

[nyngwang]

@sisrfeng Thanks for your kindness :) I resolved it days ago. You're right, the configuration isn't the point. I used the default config provided by nvim-dap-python for debugpy. (No more problems on this thread!)

[sisrfeng]

he configuration isn't the point. I used the default config provided by nvim-dap-python for debugpy. (No more problems on this thread!)

You this repo can work with gpu now ?! I tried but failed some days ago.