Incorrect network illustration

[chenfucn]

beginner_source/blitz/neural_networks_tutorial.py

The picture showing an example conv net is inconsistent with the example network in the code below. Specifically, a 32x32 image after a 3x3 conv filter with no padding should output a 30x30 channel, not a 28x28 one. After 2x2 max pool -> 15x15, another 3x3 conv -> 13x13, 2x2 max pool -> 6x6

[jimmy1411]

Only change required here is updating the kernel size of the conv layers from 3 to 5

[holly1238]

This issue was fixed and merged in #1257.