-
We will use this issue to track the results of the different network architectures and training methodologies.
Add a comment in this issue when you are working on one of them (I'll write `in progre…
-
The idea comes from keras: https://github.com/fchollet/keras/tree/master/keras/backend
Some functions have the same functional, but 2 backends can return different format of results. However, it is n…
-
Hey,
First of all thanks a lot for this. I was wondering whether there is an easy way to make the gradient flipping work in Keras. [Someone](https://github.com/fchollet/keras/issues/3119) has done…
-
For the MNIST implementation, you have the last layer of the network with 10 outputs (one for each class).
How would that work with signature-embedding, where there isn't a explicit number of class…
-
As already suggested in #162, it would be beneficial to split the very large file `returnn/tf/layers/basic.py` into smaller files with the same sorting as currently in the documentation. As splitting …
-
Hi,
Was wondering - for each element in the batch, does the current algorithm automatically parallelise? I have an RTX3090 (with 24 GiB) and I run out of memory instantly for a sequence anything lo…
-
I am proposing to be able to include a 3D deformable convolution within keras-cv
## Background
A recent paper has used deformable convolutions as a key building block in order to have a CNN be sta…
-
Hi,
I have a set of training data, and for each of those examples I want to "learn embeddings on the fly", such is done in some NLP models for example. I.e., have a bank of parameters for training ex…
-
There exist numerical instability in **objectives.categorical_crossentropy** function, which cause gradient vanishing right after the first training batch.
I suggest adding $\epsilon$ to prevent the …
-
Hello,
I've been working on reconstructing the MNIST neural net found [here](http://blog.aloni.org/posts/backprop-with-tensorflow/) using the TensorFlowSharp 1.3.0 nuget package in VisualStudio 201…