Hey, pardon for posting a similar issue twice, but I think this matter needs to some attention.
There are no proper working tensor-train RNNs for the latest tf/keras which is very odd to me. It would be fantastic to process sequences with this kind of power and speed.
As this is the most popular and active tensor library maybe you guys could give some serious thought on this matter?
Best regards
Martin
Hey, pardon for posting a similar issue twice, but I think this matter needs to some attention.
There are no proper working tensor-train RNNs for the latest tf/keras which is very odd to me. It would be fantastic to process sequences with this kind of power and speed.
As this is the most popular and active tensor library maybe you guys could give some serious thought on this matter?
Best regards
Martin