Add introspection tools, fix setup configs - Githubissues

OverLordGoldDragon / see-rnn

RNN and general weights, gradients, & activations visualization in Keras & TensorFlow

MIT License

179 stars 21 forks source link

Add introspection tools, fix setup configs #15

Closed OverLordGoldDragon closed 4 years ago

OverLordGoldDragon commented 4 years ago

Breaking:

get_layer_outputs -> get_outputs
get_layer_gradients -> get_gradients

Bug fixes:

codecs.open -> open, as former's formatting of README.md in setup.py was overflowing package build metadata
TF 2.0 / 2.1 Graph: use K_eval instead of K.get_value or .get_weights to get layer weights

Misc:

Created _backend.py to move repeated declarations and env vars to
Added K_eval to utils.py (see bugfix bullet 2)
Improved docstring for features_hist_v2
Added classifiers in setup.py