Chapter 7- issue in loading pre trained model

[xinformatics]

While following the textbook and the steps on Github on the cell counting task, when I run

model.restore()

it throws the following error:

INFO:tensorflow:Restoring parameters from model/model-7700
and
NotFoundError: 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:

2 root error(s) found.
(0) Not found: Key Conv2D_3/conv2d_4/bias not found in checkpoint
[[node save_3/RestoreV2 (defined at /tensorflow-1.15.2/python3.6/tensorflow_core/python/framework/ops.py:1748) ]]
(1) Not found: Key Conv2D_3/conv2d_4/bias not found in checkpoint
[[node save_3/RestoreV2 (defined at /tensorflow-1.15.2/python3.6/tensorflow_core/python/framework/ops.py:1748) ]]
[[save_3/RestoreV2/_101]]
0 successful operations.
0 derived errors ignored.

[rbharath]

Hmm, I think we've updated the Github code to TF 2.2 (CC @peastman) and the latest DeepChem HEAD. Could you try rerunning this code on the latest DeepChem/TF 2.2 (pip install --pre deepchem)

[peastman]

Was that loading the pretrained model? I don't think it will work with the current DeepChem release. As part of releasing DeepChem 2.4, we should create new versions of the pretrained models for the book.

[rbharath]

+1 to creating new versions of pretrained models!

[ChinmayRout9040895625]

Thanks @xinformatics for reporting this! This is a very common issue when restoring TensorFlow 1.x checkpoints.

The Problem: The error Key Conv2D_3/conv2d_4/bias not found means the current Python code defining your model's architecture does not exactly match the architecture that was used to save the model-7700 checkpoint. Even a small change (like renaming a layer or changing a scope) will break the restore operation.

Solution: Please ensure the file that defines your convolutional neural network architecture (the graph) is an exact, character-for-character match of the code provided in the official GitHub repository for that chapter. If you've modified the model code at all, please revert it to the original version and try model.restore() again.