An error message is displayed in the background (not in the jupyter notebook but in the terminal jupyter notebook was called from):
E tensorflow/core/grappler/optimizers/meta_optimizer.cc:502] layout failed: Invalid argument: Unsupported tensor size: 1
This does not seem to cause noticeable errors in the ongoing jupyter session.
Context and Steps to Reproduce:
Running the file n2v/examples/2D/denoising2D_SEM/01_training.ipynb, from a fresh clone and the conda environment with Tensorflow-gpu 1.14.0, Keras 2.2.4, n2v 0.1.9.
The issue happens during:
history = model.train(X, X_val)
Right after the first epoch is completed, and is visible in the terminal.
Alternatively, running the same code in a .py file , the issue can also be seen right after the first epoch.
Again, this does not prevent the training to continue and the results of the resulting model look okay when running the file/n2v/examples/2D/denoising2D_SEM/02_prediction.ipynb
Additional Information:
The issue does not appear when running similar code using normal csbdeep models in the same environment. This is the reason why I report this here.
I open this issue to confirm whether this could this be something affecting in some way the results or performance of the library.
Behaviour
An error message is displayed in the background (not in the jupyter notebook but in the terminal jupyter notebook was called from):
E tensorflow/core/grappler/optimizers/meta_optimizer.cc:502] layout failed: Invalid argument: Unsupported tensor size: 1This does not seem to cause noticeable errors in the ongoing jupyter session.Context and Steps to Reproduce:
Running the file n2v/examples/2D/denoising2D_SEM/01_training.ipynb, from a fresh clone and the conda environment with Tensorflow-gpu 1.14.0, Keras 2.2.4, n2v 0.1.9. The issue happens during:
history = model.train(X, X_val)Right after the first epoch is completed, and is visible in the terminal.Alternatively, running the same code in a .py file , the issue can also be seen right after the first epoch.
Again, this does not prevent the training to continue and the results of the resulting model look okay when running the file/n2v/examples/2D/denoising2D_SEM/02_prediction.ipynb
Additional Information:
The issue does not appear when running similar code using normal csbdeep models in the same environment. This is the reason why I report this here. I open this issue to confirm whether this could this be something affecting in some way the results or performance of the library.