Integration of Python and Keras (in Anaconda) in to Knime for MAC

2018-03-26 09:59:00.481670: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA

This warning message indicates that you would reach better performance if you compiled TensorFlow manually for your specific processor (instead of just installing it via conda), which is not a trivial thing to do, so you can just ignore that warning.

/Users/tracy/anaconda3/envs/py35_knime/lib/python3.5/site-packages/keras/models.py:252: UserWarning: No training configuration found in save file: the model was not compiled. Compile it manually.
warnings.warn('No training configuration found in save file: ’

This means that Keras did not detect a training configuration when it loaded the network. That’s okay, we’ll compile it (i.e. set a training configuration) manually in the learner node.

ERROR PythonKernel Exception ignored in: <bound method BaseSession.del of <tensorflow.python.client.session.Session object at 0x115736470>>
ERROR PythonKernel TypeError: ‘NoneType’ object is not callable

This is some weird error message issued by TensorFlow (bug report) which does not affect the correct execution of the network. So you can also ignore that.

It’s fairly common for a deep learning training process to take hours or days, especially when you are running on CPU. You can try to decrease the number of epochs or the number of training samples if you want to come to a result in less time, but the performance of the trained network will suffer.

You should be able to run the node in a normal manner outside the dialog (i.e. right click on the node and execute). However, the Python nodes do not report progress during execution of the script (for two hours in your case), so the node might look like it froze which is not the case.