Jana-Z
commented
2 years ago +1
I can run --eval-only on the .pkl files from the MODEL_ZOO.md. But if I train my own model it saves the weights to a .pth file and when trying to evaluate using these weights I get the following warning messages:
WARNING [07/16 15:49:00 mask2former.modeling.meta_arch.mask_former_head]: Weight format of MaskFormerHead have changed! Please upgrade your models. Applying automatic conversion now ...
WARNING [07/16 15:49:00 fvcore.common.checkpoint]: Some model parameters or buffers are not found in the checkpoint:
sem_seg_head.pixel_decoder.adapter_1.norm.{bias, weight}
sem_seg_head.pixel_decoder.adapter_1.weight
[...]
WARNING [07/16 15:49:00 fvcore.common.checkpoint]: The checkpoint state_dict contains keys that are not used by the model:
sem_seg_head.pixel_decoder.pixel_decoder.adapter_1.norm.{bias, weight}
sem_seg_head.pixel_decoder.pixel_decoder.adapter_1.weightMoreover, trying to load in the saved model_final.pth in a seperate jupyter notebook causes a PytorchStreamReader failed reading zip archive: failed finding central directory error.
Same error as #95
GuHuangAI
sushilkhadkaanon
ShijieVVu
Hello! i have truble while re-evaluating trained model.
if i reload saved weights to evaluate it using --eval-only option after training them by train_net_video.py, then error occured because the parameter name is different.
how can i solve this problem?
i captured error log.
[06/21 22:58:03] fvcore.common.checkpoint WARNING: Some model parameters or buffers are not found in the checkpoint: [34msem_seg_head.pixel_decoder.adapter_1.norm.{bias, weight}[0m [34msem_seg_head.pixel_decoder.adapter_1.weight[0m [34msem_seg_head.pixel_decoder.input_proj.0.0.{bias, weight}[0m ...
[06/21 22:58:03] fvcore.common.checkpoint WARNING: The checkpoint state_dict contains keys that are not used by the model: [35msem_seg_head.pixel_decoder.pixel_decoder.adapter_1.norm.{bias, weight}[0m [35msem_seg_head.pixel_decoder.pixel_decoder.adapter_1.weight[0m [35msem_seg_head.pixel_decoder.pixel_decoder.input_proj.0.0.{bias, weight}[0m ...