search

HYPJUDY / Decouple-SSAD

Decoupling Localization and Classification in Single Shot Temporal Action Detection
https://arxiv.org/abs/1904.07442
MIT License
96 stars 19 forks source link

Restoring from checkpoint failed #24

Open mrlihellohorld opened 3 years ago

mrlihellohorld commented 3 years ago

NotFoundError (see above for traceback): Restoring from checkpoint failed. This is most likely due to a Variable name or other graph key that is missing from the checkpoint. Please ensure that you have not altered the graph expected based on the checkpoint. Original error:

Key biClsReg_predict_layerAL1ProposalBranch/conv1d/bias not found in checkpoint [[node save/RestoreV2 (defined at decouple_ssad.py:269) ]] [[node save/RestoreV2 (defined at decouple_ssad.py:269) ]]

HYPJUDY commented 3 years ago

Hi, I haven't met this error. But I found a related issue and a solution https://github.com/tensorflow/models/issues/5003#issuecomment-485274023:

A "Key Conv/biases not found in checkpoint" error also occurs if you set your model_dir/output_path flag to point to a directory that has already been used to train a model, in which a checkpoint file has been generated.

To fix this, either choose another directory or delete the checkpoint, graph.pbtxt, events.out and model.ckpt files. Deleting the three latter ones will in some cases prevent other conflicts, but is not necessary.

Maybe you could try whether this solution could solve your problem.

mrlihellohorld commented 3 years ago

Hi, I haven't met this error. But I found a related issue and a solution tensorflow/models#5003 (comment):

A "Key Conv/biases not found in checkpoint" error also occurs if you set your model_dir/output_path flag to point to a directory that has already been used to train a model, in which a checkpoint file has been generated. To fix this, either choose another directory or delete the checkpoint, graph.pbtxt, events.out and model.ckpt files. Deleting the three latter ones will in some cases prevent other conflicts, but is not necessary.

Maybe you could try whether this solution could solve your problem.

thank u very much！ I have solved the issue, Just because I called the wrong model. But I have an other question,could u help me？the question is how to extract thumous dataset's feature through pytorch instead of using caffe. looking forward to your reply

HYPJUDY commented 3 years ago

Glad you solved your problem. About the pytorch version to extract feature, I do not have such codes.