Fix examples/mnist/mnist*.py

[ndronen]

Fixes # 2

Description:

The scripts wouldn't even run because

• Someone split Accuracy into BinaryAccuracy and CategoricalAccuracy without updating the client code!
• On a machine with a GPU, the model wasn't being put to the device, but the inputs were.

Check list:

• New tests are added (if a new feature is modified)
• New doc strings: text and/or example code are in RST format
• Documentation is updated (if required)

[vfdev-5]

@ndronen thanks for the PR, however your assumption on

Someone split Accuracy into BinaryAccuracy and CategoricalAccuracy without updating the client code!

is wrong. Historically there were two accuracy classes BinaryAccuracy, CategoricalAccuracy which have been merged into a single one. Please, take a look at #275.

Probably, you experience versions problem between the example code from the master and some old library version. I just checked the mnist.py example in a fresh conda environment and on a pc with GPU:

git clone https://github.com/pytorch/ignite.git
cd ignite/example/mnist
pip install torch torchvision tqdm numpy pytorch-ignite
python mnist.py

Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
Processing...
Done!
Training Results - Epoch: 1  Avg accuracy: 0.94 Avg loss: 0.22
Validation Results - Epoch: 1  Avg accuracy: 0.94 Avg loss: 0.20
ITERATION - loss: 0.24:  99%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████ | 930/938 [00:23<00:00, 122.01it/s]

[vfdev-5]

I'll close this PR as proposed modifications are not correct. Please create an issue if you still experience any problem with running mnist examples