goodfeli / theano_exercises Goto Github PK

Hi, there,

I'm a newbie to theano and run into a stone wall. The solution says that I'll get an error but I have not. And I want to ask how I can add this line THEANO_FLAGS = "mode=FAST_COMPILE" to the code?

Sincerely,
Lerner

02 Problem with CudaNdarray in the proposed solution.

Traceback (most recent call last):
File "ex_02_detect_negative.py", line 63, in
f(0.)
File "/home/danielcanelhas/workspace/Theano/theano/compile/function_module.py", line 588, in call
outputs = self.fn()
File "/home/danielcanelhas/workspace/Theano/theano/gof/link.py", line 761, in f
raise_with_op(node, thunk)
File "/home/danielcanelhas/workspace/Theano/theano/gof/link.py", line 759, in f
wrapper(i, node, _thunks)
File "/home/danielcanelhas/workspace/Theano/theano/gof/link.py", line 774, in wrapper
f(_args)
File "ex_02_detect_negative.py", line 41, in neg_check
do_check_on(x, node, fn)
File "ex_02_detect_negative.py", line 32, in do_check_on
if var.min() < 0:
AttributeError: 'CudaNdarray' object has no attribute 'min'
Apply node that caused the error: GpuFromHost(x)
Inputs types: [TensorType(float32, scalar)]
Inputs shapes: ['No shapes']
Inputs strides: ['No strides']
Inputs scalar values: ['not scalar']

HINT: Re-running with most Theano optimization disabled could give you a back-traces when this node was created. This can be done with by setting the Theano flags optimizer=fast_compile
HINT: Use the Theano flag 'exception_verbosity=high' for a debugprint of this apply node.

03 problem with GpuSoftmax in the proposed solution.

workarounds may be to compare the fgraph as a string containing "Softmax", though I'm not sure how much you like it:

s = str(f.maker.fgraph)
if "Softmax" in s:
    return True
return False

alternatively :

import theano.sandbox.cuda as CUDA
...
if isinstance(app.op, T.nnet.Softmax) or isinstance(app.op, CUDA.nnet.GpuSoftmax):
    return True
return False

Not really an issue, but rather a question.

I don't see the difference between both cases in the 03_energy.py example:

1. taking V and H as constants
2. differentianting over them

If I comment out the _ElemwiseNoGradient class and use T.grad(energy(W,V,H).mean(),W) in grad_expected_energy I still get correct answer.

On the other hand I get an error if I switch mode to FAST_COMPILE (both cases).

Error:
MethodNotDefined: ('perform', <class 'theano.sandbox.rng_mrg.GPU_mrg_uniform'>, 'GPU_mrg_uniform')
Apply node that caused the error: GPU_mrg_uniform{CudaNdarrayType(float32, vector),no_inplace}(<CudaNdarrayType(float32, vector)>, TensorConstant{(1,) of 12})
Inputs types: [CudaNdarrayType(float32, vector), TensorType(int32, (True,))]
Inputs shapes: [(12,), (1,)]
Inputs strides: [(1,), (4,)]
Inputs scalar values: ['not scalar', array([12], dtype=int32)]

Backtrace when the node is created:
File "03_energy.py", line 55, in
W = rng_factory.normal(size=(nv, nh), dtype=v0.dtype)
File "/local/lib/python2.7/site-packages/theano/sandbox/rng_mrg.py", line 1328, in normal
nstreams=nstreams)
File "/local/lib/python2.7/site-packages/theano/sandbox/rng_mrg.py", line 1210, in uniform
ndim, dtype, size))
File "/local/lib/python2.7/site-packages/theano/sandbox/rng_mrg.py", line 545, in new
return op(rstate, cast(v_size, 'int32'))

HINT: Use the Theano flag 'exception_verbosity=high' for a debugprint of this apply node.

It was giving this error until I turned the optimization off

RuntimeError: Error doing inplace add
Apply node that caused the error: GpuIncSubtensor{InplaceInc;::}(GpuFromHost.0, GpuArrayConstant{1.0})
Toposort index: 1
Inputs types: [GpuArrayType(float32, (False,)), GpuArrayType(float32, ())]
Inputs shapes: [(4,), ()]
Inputs strides: [(4,), ()]
Inputs values: [gpuarray.array([ 0., 0., 0., 0.], dtype=float32), gpuarray.array(1.0, dtype=float32)]
Outputs clients: [[HostFromGpu(gpuarray)(GpuIncSubtensor{InplaceInc;::}.0)]]