ask for guidance

[DisgustedALL]

I directly used the dataset (Train) you provided for training, but this error occurs why?

[jpeumesmo]

Hello @ZyzJinQu !

First of all, sorry for the (very) late reply.

Well, for your logs I am not sure what exactly occurs, I would need more information about it. But at a first look it seems some arguments are missing, you may want to use ckp_save_path and summary_dir for example. I suggest to take a deeper look into the help.

Also, I try to reproduce the code and unfortunately due deprecated OS, I couldn't reproduce entirely the environment. That's why I took so long to answer. I have update the files under setup\ folder with newer versions of OS and dependencies.

Futhermore, during the process some bugs were found and fixed. The major one is related to a change in PyTorch that change how optimizers work in version 1.4.0 (the one we used) to newer ones (more details pytorch/pytorch#39141 (comment)).

I could make a train locallly, using cpu, with that command here:

python main.py -p train -d /workspace/l2d_train/ --ckp_save_path /workspace/ckpts/ --summary_dir /workspace/logs/

Using a docker container created with the updated files.

Let me know if I can help you with anything else.