Tobias Czempiel

``` import pickle from functools import partial ... pickle.load = partial(pickle.load, encoding="latin1") pickle.Unpickler = partial(pickle.Unpickler, encoding="latin1") ckpt = torch.load(opt.load, pickle_module=pickle) ``` worked for me with python 3

> dear @tobiascz , > > import pickle > from functools import partial > ... > pickle.load = partial(pickle.load, encoding="latin1") > pickle.Unpickler = partial(pickle.Unpickler, encoding="latin1") > ckpt = torch.load(opt.load, pickle_module=pickle)...

# load ckpt if opt.load: print(">>> loading ckpt from '{}'".format(opt.load)) ckpt = torch.load(opt.load) start_epoch = ckpt['epoch'] err_best = ckpt['err'] glob_step = ckpt['step'] lr_now = ckpt['lr'] model.load_state_dict(ckpt['state_dict']) optimizer.load_state_dict(ckpt['optimizer']) print(">>> ckpt loaded...

Your code seems to be fine. maybe try loading it with and without the pickle stuff and compare them side by side in the debugger to verify if the pickle...

@NguyenDangBinh It's hard for me to help you. The problem seems to be a very specific one. I would try to make a new conda environment with python 3.6 and...

Facebook released detectron [v2](https://github.com/facebookresearch/detectron2). This tutorial was done with detectron v1 this might cause some issues.

I am using Pytorch CLI and this is the logger and callbacks part of my config: ``` logger: class_path: lightning.pytorch.loggers.wandb.WandbLogger init_args: project: "myproject" callbacks: - class_path: lightning.pytorch.callbacks.ModelCheckpoint init_args: dirpath: "/home/Develop/project/checkpoints"...