BehnoodRasti / Unmixing_Tutorial_IEEE_IADF

Codes and data for Unmixing
Apache License 2.0
44 stars 13 forks source link

Bugs in EndNet #2

Open sankalp-25 opened 1 year ago

sankalp-25 commented 1 year ago

I have been working on Endnet Autoencoder, and there are a few bugs I encountered in the code Endnet.ipynb;

  1. TensorFlow 2.2.0 is not available from python version 3.9, so removing the specification of the TensorFlow version will not throw any error.
  2. The random function is no longer present in Scipy, it is now in Numpy library. So, in VCA(vertex component analysis) algorithm w = sp.random.rand(R, 1) should be changed to w = np.random.rand(R, 1).
  3. As we are not using TensorFlow 2.2.0, the optimizer functions cannot be accessed like previously. So, while using Adam optimizer instead of opt = tf.optimizers.Adam(learning_rate=learning_rate,beta_1=0.7), we can use opt = tf.optimizers.legacy.Adam(learning_rate=learning_rate,beta_1=0.7), which will not throw any error.
idantene commented 8 months ago

A year later, and more things are (expectedly) broken, with the current showstopper (for myself, that is):

RuntimeError: tf.data.Dataset only supports Python-style iteration in eager mode or within tf.function.