DragonComputer / DragonComputer/Cerebrum

NeuralNet: Hearing to Vision LSTM seems impossible

Open
#9 0 comments 0 reactions 0 assignees View on GitHub
enhancement invalid question wontfix
Dominant language
Python
Stars
45
Forks
20
PR merge metrics
No merged PRs in 30d

Description

`h2v_net = buildNetwork(2048, 5, 230400, hiddenclass=LSTMLayer, outputbias=False, recurrent=True)`

Even with 2048 input neuron, 5 LSTM cell, 230400 output neuron, 5 epochs takes 10 seconds to complete.

For 500 epochs training every sample takes 16 minutes. Which is not an acceptable result.

It seems Hearing to Vision LSTM is impossible with today's hardware.

*It's not even starting if there are 16384 LSTM cells.

See commit 48e29c7186a19d2cbbfb56b1aaf7335e8ed9fd02

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.