To whom it may concern: To integrate the evaluation with the neural network into the server do the following:
0. make sure that you have already installed numpy, spicy and theano
1. import the neuralnetwork.py file into your source
2. create a neuralnetwork object
3. use objects load(filename) method where filename contains the prepared weight matrix
4. use neuralnetwork.evaluate(sound) to get whoever talks, where sound is a path to the .wav file
Example:
import neuralnetwork as nn
n = nn.Net() n.load('matrix.txt') print n.evaluate(sound)
To run the the train on GPU use the following code: THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python <filename.py>