Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Go, and more
that is not a good start
With this
`class Inspiration(HybridBlock):
""" Inspiration Layer (from MSG-Net paper)
tuning the featuremap with target Gram Matrix
ref https://arxiv.org/abs/1703.06953
"""
def init(self, C, B=1, ctx=mx.cpu(0)):
super(Inspiration, self).init()
""" Inspiration Layer (from MSG-Net paper)
tuning the featuremap with target Gram Matrix
ref https://arxiv.org/abs/1703.06953
"""
def init(self, C, B=1, ctx=mx.cpu(0)):
super(Inspiration, self).init()
# B is equal to 1 or input mini_batch
self.C = C
self.B = B
self.weight = self.collect_params().get('weight', shape=(1,self.C,self.C),
init=mx.initializer.Uniform(),
allow_deferred_init=True)
self.gram = self.collect_params().get('gram', shape=(self.B,self.C,self.C),
init=mx.initializer.Uniform(),
allow_deferred_init=True,
lr_mult=0)
self.weight.initialize(ctx=ctx)
self.gram.initialize(ctx=ctx)
def setTarget(self, target):
self.gram.set_data(target)
def hybrid_forward(self, F, X, gram, weight):
P = F.batch_dot(F.broadcast_to(weight, shape=(self.gram.shape)), gram)
if not isinstance(X,symbol.Symbol):
return F.batch_dot(F.SwapAxis(P,1,2).broadcast_to((X.shape[0], self.C, self.C)), X.reshape((0,0,X.shape[2]*X.shape[3]))).reshape(X.shape)
else:
#print "Hooppla", interals
#for i in dir(interals):
# print "kk:", i
in_shapes,out_shapes,arg_shapes= X.infer_shape(self.gram.shape)
#print out_shapes
#raise Exception
#arg_shapes, out_shapes, aux_shapes = interals.infer_shape(self.gram.shape)
#print "A", arg_shapes, "O", out_shapes, "AU", aux_shapes
return F.batch_dot(F.SwapAxis(P,1,2).broadcast_to((in_shapes[0], self.C, self.C)), X.reshape((0,0,in_shapes[2]*in_shapes[3]))).reshape(in_shapes)
def __repr__(self):
return self.__class__.__name__ + '(' \
+ 'N x ' + str(self.C) + ')'`
i get the following error
thanks for taking time out to work on this !
out_shape is an empty tuple in a list
how can I use that?
In C++
you are able to hybridize your model correct ?
but the model needs to know the resolution of the input data
before you can serialise it
let me show you what I mean
this commit kind of spells it out: samhodge/incubator-mxnet@1d72d60
I need to save parts of the model and put the pieces together in C++
https://mxnet.incubator.apache.org/api/python/symbol.html#mxnet.symbol.reshape yeah maybe it is not an issue
I can just reshape the appropriate symbole
it feels like I am trying to rough out an idea at this stage
https://github.com/samhodge/incubator-mxnet/blob/master/cpp-package/include/mxnet-cpp/symbol.hpp this has most of what I need
Actually dont worry about that, I think I need to continue on with a fixed resolution model first and simply try that out with the feedforward in C++ and then build it up from there, I am a little confused about if it is possible to build the gram matrix and the inspiration network in C++, but as a start it might be worth just loading up the model and params that I have for a decent trained model.
I just worry that I am delaying a decision
so the link that you pointed to https://github.com/samhodge/incubator-mxnet/blob/master/cpp-package/include/mxnet-cpp/symbol.hpp is just the cpp package, its just another frontend like python.
width*