Tensorflow running eagerly TODO

What is the difference between setting run_eagerly while compiling a tf.keras model vs setting it after compilation vs tf.config.experimental_run_functions_eagerly(True)?

Tensorflow saving models blues

As of Di 03 Mär 2020 17:11:47 CETA, it seems saving weights both in .h5 format and tf format is broken. For TF format, [TF2.0] Bug when saving weights with custom layers · Issue #26811 · tensorflow/tensorflow, you should use .h5; For .h5, Tensorflow 2.x seems to use the same names of Variables if they are unnamed, which creates conflicts when saving. The fix is to name all the variables: Custom layer weights all have the same name by default · Issue #36650 · tensorflow/tensorflow

Zinc for common colds

This paper: Zinc for the common cold—not if, but when says that it’s conclusive evidence that it helps at <24h after the start and that it reduces the risk of colds. I’m not sure I understand what’s the right dose though