budach / pysster Goto Github PK

Hi,

I'm having trouble adding dense layers of with different neuron numbers. For instance, if I would like 3 hidden layers of widths 300, 100, 30 - is there a way to initialize a pysster model to handle this?

Thank you for your help!

Todo: remove the keras dependency and switch to tf.keras when Tensorflow 2.0 is released.

Hi, just spotted a small issue with the documentation of utils.annotate_structures - as well as H,I,M,S forgi can output F (fiveprime) and T (threeprime) as labels, it might be good if this was mentioned to avoid confusion.

Simply replacing all occurrences of from keras with from tensorflow.python.keras works like a charm - no crashes - BUT: the loss on my data sets doesn't improve over time anymore. Other users observe similar behavior (tensorflow/tensorflow#15026, tensorflow/tensorflow#15831).

I will try again with a future tensorflow version. Having one less dependency is always nice.

Hi I was wondering if adding a tensorboard option is planned.

For each kernel, there are a lot of indidivual output files to look at right now (sequence logo, violin plots, positional enrichment plots). I guess it would be convenient to generate an html summary showing all plots to get a quick overview by just scrolling through that single html document.

Hello, may I ask if it is possible to provide the structure strings for the data sets in /tutorials/rbp/data/ obtained via ViennaRNA python bindings (in the measure_rbp_pysster.py)?

I tried to do this by running the measure_rbp_pysster.py file but I have trouble installing the ViennaRNA and cannot resolve the installation problem.

Thank you very much.

When training two pysster Model objects with the same seed, model weights are not equal anymore with keras 2.3.0. Respective tests have been disabled for now. Might not need a fix, since we want to switch to tf.keras anyway.