I have tried numerous example of tensorflow models but run to problems with Keras, datasets, MNIST, CIFAR10, etc. Does anybody know of an actual example of a tensorflow model that actually works?? I have probably tried 10 so called simple tensorflow models and have gotten numerous issues.
BTW, I NEVER had this much trouble with Pytorch so I proffer that I am not an idiot
(Successfully build CNN, RNN, etc in pytorch)
Hi @Ted_Vojnovich, I was able to train a model by loading the mnist dataset from keras datasets. Please refer to this gist for code example. Thank You.
BBBBZZZZ want a lifeline!!! Pytorch is not this hard!!!
Does not work…reads garbage
import tensorflow as tf
import keras
import numpy as np
mnist=keras.datasets.mnist.load_data()
(x_train,y_train),(x_test,y_test)=mnist
BTW…pytorch works fine
#Width, Color Channels)
transform = transforms.ToTensor()
print(“transform loaded”)
#Train Data
train_data = datasets.MNIST(root=‘/mnist_csv’, train=True, download=True,
transform=transform)
test_data = datasets.MNIST(root=‘/mnist_csv’, train=False, download=True,
transform=transform)
update: Both MNIST and CIDAR10 do NOT work for no fewer than 10 examples I found
so I did prove that CSV classification in Tensorflow works (with a few minor hiccups…but enough to prove reading the data and throwing it a a model works fine).
From what I can tell, every example of MNIST or CIDAR10 must not document or must assume something needed. All I give is essentiall a giant memory dump of cryptic characters to the terminal. If Google ever decides write good code (android memory leaks, lack of discipline in defining functions in sheets, bringing back the implicit context sensitve interface for sketchup…and now with tensorflow), Google could make a legitimate argument about superiority 
Hi @Ted_Vojnovich, Could you please elaborate more on this like garbage in the scenes. Thank You.