# another, and what about __hash__ ? So the unpickled version doesn't
# equal the pickled version, and the cmodule cache is not happy with
# the situation. The old back-end have this same comment and use the
# same mechanism.
def GpuArray_unpickler(npa, ctx_name):
if config.experimental.unpickle_gpu_on_cpu:
# directly return numpy array
warnings.warn(
"config.experimental.unpickle_gpu_on_cpu is set to True. "
"Unpickling GpuArray as numpy.ndarray")
return npa
elif pygpu:
ctx = get_context(ctx_name)
return pygpu.gpuarray.array(npa, copy=True, context=ctx)
else:
raise ImportError("pygpu not found. Cannot unpickle GpuArray")
copyreg.constructor(GpuArray_unpickler)
def GpuArray_pickler(cnda):
ctx_name = _name_for_ctx(cnda.context)
return (GpuArray_unpickler, (np.asarray(cnda), ctx_name))
# In case pygpu is not imported.
if pygpu is not None:
copyreg.pickle(pygpu.gpuarray.GpuArray, GpuArray_pickler,
GpuArray_unpickler)
# THIS WORKS But CudaNdarray instances don't compare equal to one
# another, and what about __hash__ ? So the unpickled version doesn't
# equal the pickled version, and the cmodule cache is not happy with
# the situation.
def CudaNdarray_unpickler(npa):
if (config.experimental.unpickle_gpu_on_cpu and config.device == 'cpu'):
# directly return numpy array
warnings.warn(
"config.experimental.unpickle_gpu_on_cpu is set to True. Unpickling CudaNdarray as numpy.ndarray"
)
return npa
elif cuda:
return cuda.CudaNdarray(npa)
else:
raise ImportError("Cuda not found. Cannot unpickle CudaNdarray")
copyreg.constructor(CudaNdarray_unpickler)
def CudaNdarray_pickler(cnda):
return (CudaNdarray_unpickler, (numpy.asarray(cnda), ))
# In case cuda is not imported.
if cuda is not None:
copyreg.pickle(cuda.CudaNdarray, CudaNdarray_pickler,
CudaNdarray_unpickler)
version=3,
)
# THIS WORKS But CudaNdarray instances don't compare equal to one
# another, and what about __hash__ ? So the unpickled version doesn't
# equal the pickled version, and the cmodule cache is not happy with
# the situation.
def CudaNdarray_unpickler(npa):
if config.experimental.unpickle_gpu_on_cpu:
# directly return numpy array
warnings.warn("config.experimental.unpickle_gpu_on_cpu is set to True. Unpickling CudaNdarray as numpy.ndarray")
return npa
elif cuda:
return cuda.CudaNdarray(npa)
else:
raise ImportError("Cuda not found. Cannot unpickle CudaNdarray")
copyreg.constructor(CudaNdarray_unpickler)
def CudaNdarray_pickler(cnda):
return (CudaNdarray_unpickler, (numpy.asarray(cnda),))
# In case cuda is not imported.
if cuda is not None:
copyreg.pickle(cuda.CudaNdarray, CudaNdarray_pickler, CudaNdarray_unpickler)
# THIS WORKS But GpuArray instances don't compare equal to one
# another, and what about __hash__ ? So the unpickled version doesn't
# equal the pickled version, and the cmodule cache is not happy with
# the situation. The old back-end have this same comment and use the
# same mechanism.
def GpuArray_unpickler(npa, ctx_name):
if config.experimental.unpickle_gpu_on_cpu:
# directly return numpy array
warnings.warn(
"config.experimental.unpickle_gpu_on_cpu is set to True. "
"Unpickling GpuArray as numpy.ndarray")
return npa
elif pygpu:
ctx = get_context(ctx_name)
return pygpu.gpuarray.array(npa, copy=True, context=ctx)
else:
raise ImportError("pygpu not found. Cannot unpickle GpuArray")
copyreg.constructor(GpuArray_unpickler)
def GpuArray_pickler(cnda):
ctx_name = _name_for_ctx(cnda.context)
return (GpuArray_unpickler, (numpy.asarray(cnda), ctx_name))
# In case pygpu is not imported.
if pygpu is not None:
copyreg.pickle(pygpu.gpuarray.GpuArray,
GpuArray_pickler,
GpuArray_unpickler)
if real_func is None:
method = lambda: None
else:
method = getattr(classifier_cls, real_func)
if hasattr(method, 'im_self') and getattr(method, 'im_self'):
types.MethodType(method, proxy_class)
elif not hasattr(method, 'im_self'):
method = staticmethod(method)
setattr(proxy_class, proxy_func, method)
return proxy_class
copyreg.constructor(_create_proxy)
copyreg.pickle(_dynamic_proxy_class, _dynamic_proxy_class.__reduce__, _create_proxy)
class ProxyClassifierFactory(object):
def __init__(self, classifier_cls, learn_func=None, predict_func=None, weights_func=None):
learn_func, predict_func, weights_func = \
ProxyClassifierFactory.__find_funcs(classifier_cls, learn_func, predict_func, weights_func)
self.__proxyclass = _create_proxy(
classifier_cls,
learn_func,
predict_func,
weights_func
)
