CODEWITHSUNDEEP
pythontensorflow
Troper
Troper

Reputation: 178

How to feed python lists into Tensorflow

I've processed some data for usage with tensorflow into two python lists. Unfortunately, Tensorflow does not support Python lists as an input for fitting a model. A short and simple code snippet of what my current implementation looks like follows below:

data = loadData()
pythonListInputData = data[0]
pythonListInputLabels = data[1]
model = prepModel() #note that this uses keras
model.fit(pythonListInputData, pythonListInputLabels) #Leaving out my configuration settings here for simplicities sake

This results in an error:

ValueError: Please provide as model inputs either a single array or a list of arrays.

I get this error when pythonListInputData is either a list of lists or a list of arrays.

I've read through the Tensorflow tutorials and documentation but am struggling to find any useful info that makes this work.

Edit: Data structure is as follows:

pythonListInputData = data[0]

is a list of lists of integers e.g. [[234, 1, 4], [245, 2, 5], [123, 5, 11], ...]

I have also tried an alternative which is constructed as follows

pythonListInputData = []
for entry in data[0]:
    pythonListInputData.append(array.array('I', entry))

where data[0] is the format mentioned above.

pythonListInputLabels = data[1]

is a list of integers e.g. [1, 4, 2, ...]

Upvotes: 2

Views: 1269

Answers (1)

Peter Barrett Bryan
Peter Barrett Bryan

Reputation: 675

Sam! I'm posting here in case the answer is useful to folks in the future. We came to the answer in comments.

The issue is that the keras/ tensorflow model is expecting a numpy array np.array(listName)

That should fix your issue! Cheers!

Upvotes: 3

Related Questions