bean_classifier.py

from __future__ import print_function

import numpy as np
import tflearn

from tflearn.data_utils import load_csv
data, labels = load_csv('extracted_histograms.csv', target_column=0, n_classes=2)

def preprocess(data, columns_to_ignore):
    # Sort by descending id and delete columns
    for id in sorted(columns_to_ignore, reverse=True):
        [r.pop(id) for r in data]
    return np.array(data, dtype=np.float32)

to_ignore=[0] # file name
data = preprocess(data, to_ignore)
# print(len(data[0])) # 30 elements, as expected

# Build neural network
net = tflearn.input_data(shape=[None,30])
net = tflearn.fully_connected(net, 32)
net = tflearn.fully_connected(net, 32)
net = tflearn.fully_connected(net, 2, activation='softmax')
net = tflearn.regression(net)

# Define model
model = tflearn.DNN(net)
# Start training (apply gradient descent algorithm)
model.fit(data, labels, n_epoch=10, validation_set=0.1, show_metric=True)

Oh, JFC, nice error message, tflearn!

Turns out it was in my way of loading data:

categorical_labels: bool. If True, labels are returned as binary vectors (to be used with 'categorical_crossentropy').

If I don't set that to true, it throws an error on the size of the input. Clear as mud.