With "optimizer=None", no graph optimizations are applied, in particular:
- inplace operations are not inserted, and
- you are going to use an extremely inefficient implementation of convolution
I would advise "optimizer=fast_compile" or "optimizer=fast_run" (the
default) instead.
On Fri, Nov 11, 2016, Alberto Olivero wrote:
> I use Ubuntu 16.04
> Theano 0.8
> Lasagne 0.2dev2
> nolearn 0.61dev0
>
> in the file .theanorc
> [global]
> floatX=float32
> device=cpu
> optimizer=None
>
> I set to run on the CPU just to simplify and because I have GPU with only
> 1Gb that is not enought for this model.
> The 4Gb of RAM on the main board appear fully used during the training.
>
> My Python come simply imported the VGG convolutional neural net availabe in
> internet written in Lasagne that I rewrote in nolearn.
> At the time when the fit function is working
>
> cnn.fit(train224,target)
>
> I receive the following error message after the description of the net
> provided by nolearn.
>
>
> # Neural Network with 119586826 learnable parameters
>
> ## Layer information
>
> name size total cap.Y cap.X cov.Y cov.X
> ---------- ----------- ------- ------- ------- ------- -------
> input 3x224x224 150528 100.00 100.00 100.00 100.00
> conv1_1 64x224x224 3211264 100.00 100.00 1.34 1.34
> conv1_2 64x224x224 3211264 60.00 60.00 2.23 2.23
> pool1 64x112x112 802816 60.00 60.00 2.23 2.23
> conv2_1 128x112x112 1605632 66.67 66.67 4.02 4.02
> conv2_2 128x112x112 1605632 46.15 46.15 5.80 5.80
> pool2 128x56x56 401408 46.15 46.15 5.80 5.80
> conv3_1 256x56x56 802816 57.14 57.14 9.38 9.38
> conv3_2 256x56x56 802816 41.38 41.38 12.95 12.95
> conv3_3 256x56x56 802816 32.43 32.43 16.52 16.52
> pool3 256x28x28 200704 32.43 32.43 16.52 16.52
> conv4_1 512x28x28 401408 45.28 45.28 23.66 23.66
> conv4_2 512x28x28 401408 34.78 34.78 30.80 30.80
> conv4_3 512x28x28 401408 28.24 28.24 37.95 37.95
> pool4 512x14x14 100352 28.24 28.24 37.95 37.95
> conv5_1 512x14x14 100352 41.03 41.03 52.23 52.23
> conv5_2 512x14x14 100352 32.21 32.21 66.52 66.52
> conv5_3 512x14x14 100352 26.52 26.52 80.80 80.80
> pool5 512x7x7 25088 26.52 26.52 80.80 80.80
> fc6 4096 4096 100.00 100.00 100.00 100.00
> fc6dropout 4096 4096 100.00 100.00 100.00 100.00
> fc7 4096 4096 100.00 100.00 100.00 100.00
> fc7dropout 4096 4096 100.00 100.00 100.00 100.00
> fc8 10 10 100.00 100.00 100.00 100.00
>
> Explanation
> X, Y: image dimensions
> cap.: learning capacity
> cov.: coverage of image
> magenta: capacity too low (<1/6)
> cyan: image coverage too high (>100%)
> red: capacity too low and coverage too high
>
>
> Traceback (most recent call last):
> File "/home/alberto/pycharm-community-2016.2.3/helpers/pydev/pydevd.py",
> line 1580, in <module>
> globals = debugger.run(setup['file'], None, None, is_module)
> File "/home/alberto/pycharm-community-2016.2.3/helpers/pydev/pydevd.py",
> line 964, in run
> pydev_imports.execfile(file, globals, locals) # execute the script
> File "/home/alberto/Scrivania/deep_learn/mnist_example/kagle
> dataset_2_VGG.py", line 248, in <module>
> cnn.fit(train224,target) # train the CNN model for n epochs
> File
> "/home/alberto/.local/lib/python2.7/site-packages/nolearn/lasagne/base.py",
> line 674, in fit
> self.train_loop(X, y, epochs=epochs)
> File
> "/home/alberto/.local/lib/python2.7/site-packages/nolearn/lasagne/base.py",
> line 737, in train_loop
> self.apply_batch_func(self.train_iter_, Xb, yb))
> File
> "/home/alberto/.local/lib/python2.7/site-packages/nolearn/lasagne/base.py",
> line 828, in apply_batch_func
> return func(Xb) if yb is None else func(Xb, yb)
> File
> "/home/alberto/.local/lib/python2.7/site-packages/theano/compile/function_module.py"
> , line 871, in __call__
> storage_map=getattr(self.fn, 'storage_map', None))
> File "/home/alberto/.local/lib/python2.7/site-packages/theano/gof/link.py"
> , line 314, in raise_with_op
> reraise(exc_type, exc_value, exc_trace)
> File
> "/home/alberto/.local/lib/python2.7/site-packages/theano/compile/function_module.py"
> , line 859, in __call__
> outputs = self.fn()
> MemoryError:
> Apply node that caused the error: Elemwise{add,no_inplace}(AbstractConv2d{
> border_mode=(1, 1), subsample=(1, 1), filter_flip=False, imshp=(None, 3, 224
> , 224), kshp=(64, 3, 3, 3)}.0, DimShuffle{x,0,x,x}.0)
> Toposort index: 65
> Inputs types: [TensorType(float32, 4D), TensorType(float32, (True, False,
> True, True))]
> Inputs shapes: [(128, 64, 224, 224), (1, 64, 1, 1)]
> Inputs strides: [(12845056, 200704, 896, 4), (256, 4, 4, 4)]
> Inputs values: ['not shown', 'not shown']
> Outputs clients: [[Elemwise{abs_,no_inplace}(Elemwise{add,no_inplace}.0),
> Elemwise{add,no_inplace}(Elemwise{add,no_inplace}.0, Elemwise{abs_,
> no_inplace}.0)]]
>
> Backtrace when the node is created(use Theano flag traceback.limit=N to
> make it longer):
> File "/home/alberto/pycharm-community-2016.2.3/helpers/pydev/pydevd.py",
> line 1580, in <module>
> globals = debugger.run(setup['file'], None, None, is_module)
> File "/home/alberto/pycharm-community-2016.2.3/helpers/pydev/pydevd.py",
> line 964, in run
> pydev_imports.execfile(file, globals, locals) # execute the script
> File "/home/alberto/Scrivania/deep_learn/mnist_example/kagle
> dataset_2_VGG.py", line 241, in <module>
> cnn.initialize()
> File
> "/home/alberto/.local/lib/python2.7/site-packages/nolearn/lasagne/base.py",
> line 479, in initialize
> self.y_tensor_type,
> File
> "/home/alberto/.local/lib/python2.7/site-packages/nolearn/lasagne/base.py",
> line 602, in _create_iter_funcs
> layers, target=y_batch, **objective_kw)
> File
> "/home/alberto/.local/lib/python2.7/site-packages/nolearn/lasagne/base.py",
> line 189, in objective
> output_layer, deterministic=deterministic, **get_output_kw)
> File
> "/home/alberto/.local/lib/python2.7/site-packages/lasagne/layers/helper.py",
> line 191, in get_output
> all_outputs[layer] = layer.get_output_for(layer_inputs, **kwargs)
> File
> "/home/alberto/.local/lib/python2.7/site-packages/lasagne/layers/conv.py",
> line 337, in get_output_for
> activation = conved + self.b.dimshuffle(('x', 0) + ('x',) * self.n)
>
> HINT: Use the Theano flag 'exception_verbosity=high' for a debugprint and
> storage map footprint of this apply node.
> Backend TkAgg is interactive backend. Turning interactive mode on.
> We've got an error while stopping in post-mortem: <type 'exceptions.
> KeyboardInterrupt'>
>
>
> Process finished with exit code 1
>
>
> --
>
> ---
> 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 [email protected].
> For more options, visit https://groups.google.com/d/optout.
--
Pascal
--
---
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 [email protected].
For more options, visit https://groups.google.com/d/optout.
