CenterSnap
CenterSnap copied to clipboard
Error while trying to run finetune on real data
I'm trying to run the finetuning code:
./runner.sh net_train.py @configs/net_config_real_resume.txt --checkpoint \path\to\best\checkpoint
But I get this error: Can't pickle <class 'zstd.ZstdError'>
This happens when the code tries to load the checkpoint in common.py script: torch.load(hparams.checkpoint, map_localization='cpu')['state_dict']
zstandard and all the other dependencies are correctly installed on my machine, and the checkpoint path is correct. Seems there is an issue in parsing the checkpoint, but I don't know what it is. Any suggestion or help would be appreciated.