Move KnetNLPModel example from JOSS #69
Conversation
paraynaud
commented
Jun 30, 2023
paraynaud
commented
- fix source code, remove Chain effectively fix Remove KnetNLPModels.Chain #64
paraynaud
force-pushed
the
pr-mv-JOSS-knetexample
branch
from
e0a374e
to
304b866
Compare
docs/src/LeNetTraining.md
Outdated
| @@ -0,0 +1,123 @@ | |||
| # Training a LeNet architecture with JSO optimizers | |||
There was a problem hiding this comment.
Can we use LeNet_training.md as a name, so we are consistence with JSO style
docs/src/LeNetTraining.md
Outdated
| ``` | ||
|
|
||
| ## MNIST dataset loading | ||
| Accordingly to LeNet architecture, we chose the MNIST dataset [@lecun-bouttou-bengio-haffner1998] from MLDataset: |
There was a problem hiding this comment.
Does it find the reference or we need to add references manually ?
There was a problem hiding this comment.
It will not find it.
Thanks for the catch, I will fix this soon.
|
|
||
| dtrn = minibatch(xtrn, ytrn, 100; xsize=(size(xtrn, 1), size(xtrn, 2), 1, :)) # training minibatch | ||
| dtst = minibatch(xtst, ytst, 100; xsize=(size(xtst, 1), size(xtst, 2), 1, :)) # test minibatch | ||
| dtrn = minibatch(xtrain, ytrain, 100; xsize=(size(xtrain, 1), size(xtrain, 2), 1, :)) # training minibatch |
There was a problem hiding this comment.
would it be better to create a variable call b_size and pass it so the reader understand the batch size is 100
There was a problem hiding this comment.
We should clarify (homogenize) the minibatch sizes.
Because, in this case, 100 indicates that each minibtach is a hundreth (1/100) of the given dataset.
For example, a MNIST training minibatch is of size 600 (60000/100)
There was a problem hiding this comment.
I am good with this, @dpo should I approve it (first review :) )
There was a problem hiding this comment.
I am good with this, @dpo should I approve it (first review :) )
