missing predict.py

[linksyncjameshwartlopez]

I wonder where is predict.py file that was mention in https://www.youtube.com/watch?v=u9FPqkuoEJ8 at 6:41 of the video

[hcchengithub]

There's none such file. But it can be done manually:

model.load(pathname) 
y, sr = librosa.load("PathName.wav",mono=True)
mfcc = librosa.feature.mfcc(y,sr) 
MFCC = np.pad(mfcc,((0,0),(0,80-len(mfcc[0]))),mode='constant',constant_values=0) 
model.predict([MFCC]) 

https://github.com/hcchengithub/tensorflow_speech_recognition_demo

[linksyncjameshwartlopez]

Should it be something like this. Im only guessing because I don't how to build up model variable. kindly correct me if im wrong.

`from future import division, print_function, absolute_import
import tflearn
import speech_data
import tensorflow as tf

learning_rate = 0.0001
training_iters = 300000 # steps
batch_size = 64

width = 20 # mfcc features
height = 80 # (max) length of utterance
classes = 10 # digits

batch = word_batch = speech_data.mfcc_batch_generator(batch_size)
X, Y = next(batch)
trainX, trainY = X, Y
testX, testY = X, Y #overfit for now

net = tflearn.input_data([None, width, height])
net = tflearn.lstm(net, 128, dropout=0.8)
net = tflearn.fully_connected(net, classes, activation='softmax')
net = tflearn.regression(net, optimizer='adam', learning_rate=learning_rate, loss='categorical_crossentropy')

col = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)
for x in col:
tf.add_to_collection(tf.GraphKeys.VARIABLES, x )

model = tflearn.DNN(net, tensorboard_verbose=0)
model.load(pathname)
y, sr = librosa.load("PathName.wav",mono=True)
mfcc = librosa.feature.mfcc(y,sr)
MFCC = np.pad(mfcc,((0,0),(0,80-len(mfcc[0]))),mode='constant',constant_values=0)
model.predict([MFCC])`

[hcchengithub]

No, you need to know how to use tflearn's model.save() and model.load() to save-restore your trained networks. You need to complete training your model, get the saved network so you don't need to do the training every time when you only want to try some predicts. They are not difficult. Just do it correctly and you can make it. My snippet assumes that you have the saved network already and start from there shows you how to do the prediction.

[linksyncjameshwartlopez]

Thanks @hcchengithub , one last question how do you instantiate model variable in your code snippet? I've tried to look at it and i arrive on this https://stackoverflow.com/questions/45099608/tflearn-issue-over-model-load?answertab=active#tab-top.

Or am i doing it wrong. Im so naive that i don't know how to load previously saved model of tflearn. Hope you can help me with this :). i'll try to dig more if i can find answers myself, it seems to be so obvious to the person who already done loading of a saved model but for a person who haven't done it yet is really vague.

[hcchengithub]

This demo is not really suitable for totally new comers. It's uncompleted ,incorrect and useless, really.
Don't waste your time here.
For general AI learning, I suggest http://www.hvass-labs.org/ or Google developers Josh Gordon's {Machine Learning} Recipes
For speech recognition, jump to Mozilla DeepSpeech.

[mohsin671]

Hi @hcchengithub , were you able to run the code? I am having this error:
ValueError: At least two variables have the same name: FullyConnected/W