Re: [theano-users] Announcement: You can training lasagne model in DIGITS now.

[Pascal Lamblin]

Just forwarding the announcement.

On 2017-10-22 07:31 AM, Xuesong Guo wrote:
> I'm happy to announce that I have made some modifies to DIGITS to
> support training lasagne model.
>
> You need install [deepstacks.lasagne
> <https://github.com/guoxuesong/deepstacks>](https://github.com/guoxuesong/deepstacks)
> and [a modified version of DIGITS
> <https://github.com/guoxuesong/DIGITS>](https://github.com/guoxuesong/DIGITS).
> You still need to install caffe, because DIGITS need it.
>
> Please download mnist dataset following [DIGITS's GettingStarted
> document
> <https://github.com/NVIDIA/DIGITS/blob/master/docs/GettingStarted.md>](https://github.com/NVIDIA/DIGITS/blob/master/docs/GettingStarted.md).
> When creating model, select Deepstacks.lasagne instead of caffe. Click
> 'customize' of LeNet, and paste the following code:
>
> ```
> |
> # Example for using raw lasagne:
>
> importdeepstacks
> importlasagne
> fromdeepstacks.framework.main import*
> importdeepstacks.framework.using_lmdb
> fromdeepstacks.macros import*
> fromdeepstacks.framework.macros import*
> defbuild_network(inputs):
>     network =inputs['image']
> if'mean'ininputs:
>
> network=lasagne.layers.ElemwiseMergeLayer((network,inputs['mean']),T.sub)
>     network =lasagne.layers.Conv2DLayer(
>             network,num_filters=20,filter_size=(5,5),
>             nonlinearity=lasagne.nonlinearities.rectify,
>             W=lasagne.init.GlorotUniform())
>     network =lasagne.layers.MaxPool2DLayer(network,pool_size=(2,2))
>     network =lasagne.layers.Conv2DLayer(
>             network,num_filters=50,filter_size=(5,5),
>             nonlinearity=lasagne.nonlinearities.rectify,
>             W=lasagne.init.GlorotUniform())
>     network =lasagne.layers.MaxPool2DLayer(network,pool_size=(2,2))
>     network =lasagne.layers.DenseLayer(
>             lasagne.layers.dropout(network,p=.5),
>             num_units=500,
>             nonlinearity=lasagne.nonlinearities.rectify)
>     network =lasagne.layers.DenseLayer(
>             lasagne.layers.dropout(network,p=.5),
>             num_units=10,
>             nonlinearity=lasagne.nonlinearities.softmax)
>
>
> res,stacks,paramlayers,errors,watchpoints=deepstacks.lasagne.build_network(network,(
> (classify,'target'),
> ),inputs)
> returnres,stacks,paramlayers+[network],errors,watchpoints
> |
>
>
> ```
>
> 'Show visualizations and statistics' and 'Classify Many' feature should
> be ok.
>
> Enjoy it.
>
> --
>
> ---
> You received this message because you are subscribed to the Google
> Groups "theano-users" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to theano-users...@googlegroups.com
> <mailto:theano-users...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Pascal Lamblin