本文整理汇总了Python中theano.sandbox.cuda.basic_ops.host_from_gpu函数的典型用法代码示例。如果您正苦于以下问题：Python host_from_gpu函数的具体用法？Python host_from_gpu怎么用？Python host_from_gpu使用的例子？那么恭喜您, 这里精选的函数代码示例或许可以为您提供帮助。
在下文中一共展示了host_from_gpu函数的15个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: local_large_sparse_targets_gpu
def local_large_sparse_targets_gpu(node):
if not isinstance(node.op, LargeSparseTargets) or theano.config.device == "cpu":
return False
if node.op.what_to_output == 0:
return [GpuLargeSparseTargets(node.op.what_to_output)(*node.inputs)]
elif node.op.what_to_output == 1:
return [host_from_gpu(GpuLargeSparseTargets(node.op.what_to_output)(*node.inputs))]
else:
out = GpuLargeSparseTargets(node.op.what_to_output)(*node.inputs)
return [out[0], host_from_gpu(out[1])]示例2: use_gpu_images2neibs
def use_gpu_images2neibs(node):
if (type(node.op) is Images2Neibs and
node.inputs[0].dtype == 'float32' and
node.op.mode in ['valid', 'wrap_centered']):
return [host_from_gpu(gpu_images2neibs(gpu_from_host(node.inputs[0]),
node.inputs[1], node.inputs[2],
mode=node.op.mode))]示例3: use_gpu_cumsum
def use_gpu_cumsum(node):
if type(node.op) is CumOp \
and node.inputs[0].dtype == 'float32' \
and node.inputs[0].owner \
and isinstance(node.inputs[0].owner.op, HostFromGpu):
if node.op.mode != 'add':
return None
axis = node.op.axis
x = node.inputs[0]
if axis is not None and x.ndim > GpuCumsum.SUPPORTED_NDIMS:
return None
x = gpu_from_host(x)
if axis is None and x.ndim > 1:
x = gpu_flatten(x)
# ``gpu_cumsum`` assume array has been flattened if needed.
if axis is None:
axis = 0
ret = host_from_gpu(GpuCumsum(axis)(x))
ret.tag.values_eq_approx = values_eq_approx_high_tol
return [ret]示例4: local_gpu_minres
def local_gpu_minres(node):
if isinstance(node.op, MinresQLP):
sw = False
for inp in node.inputs:
if inp.owner and inp.owner.op == host_from_gpu:
sw = True
if sw:
inps = node.inputs
nw_inps = []
for inp in inps:
if not isinstance(inp.type, CudaNdarrayType):
nw_inps.append(gpu_from_host(inp))
else:
nw_inps.append(inp)
new_op = node.op
new_op.gpu = 1
_new_outs = node.op(*nw_inps)
new_outs = []
for out in _new_outs:
if isinstance(out.type, CudaNdarrayType):
new_outs.append(host_from_gpu(out))
else:
new_outs.append(out)
return new_outs
else:
return False示例5: use_gpu_images2neibs
def use_gpu_images2neibs(node):
if type(node.op) is Images2Neibs:
return [
host_from_gpu(
gpu_images2neibs(gpu_from_host(node.inputs[0]), node.inputs[1], node.inputs[2], mode=node.op.mode)
)
]示例6: local_gpu_argmax
def local_gpu_argmax(node):
if type(node.op) is KArgmax:
p, = node.inputs
vals, indx, = node.outputs
if (p.dtype == vals.dtype == 'float32' and
any([i.owner and isinstance(i.owner.op, theano.sandbox.cuda.HostFromGpu) for i in node.inputs])):
gpu_op = GpuKArgmax(node.op.K)
ret_vals, ret_indx = gpu_op(gpu_from_host(p))
return [host_from_gpu(ret_vals), T.cast(host_from_gpu(ret_indx), "int32")]
if (isinstance(node.op, theano.sandbox.cuda.GpuFromHost) and
node.inputs[0].owner and type(node.inputs[0].owner.op)
is KArgmax):
multi = node.inputs[0].owner
p, = multi.inputs
vals, indx, = multi.outputs
if (p.dtype == vals.dtype == 'float32'):
gpu_op = GpuKArgmax(node.inputs[0].owner.op.K)
ret_vals, ret_indx = gpu_op(gpu_from_host(p))
return [gpu_from_host(ret_vals), gpu_from_host(ret_indx)]示例7: local_gpu_conv3d
def local_gpu_conv3d(node):
if isinstance(node.op, Conv3D):
if numpy.any([i.owner and isinstance(i.owner.op, HostFromGpu)
for i in node.inputs]):
if numpy.all([o.type.dtype == 'float32' for o in node.outputs]):
V, W, b, d = node.inputs
return [host_from_gpu(gpu_convd(as_cuda_ndarray_variable(V),
as_cuda_ndarray_variable(W),
as_cuda_ndarray_variable(b),
d))]示例8: save_data
def save_data(self, filename, data):
if type(data) != type(np.asarray([])):
data = host_from_gpu(data)
data = np.asarray(data.eval())
mult = lambda x, y: x * y
length = reduce(mult, data.shape)
data = data.reshape(length)
data = "\n".join([str(i) for i in data])
f = open(filename, "w")
f.write(data)
f.close()示例9: use_gpu_images2neibs
def use_gpu_images2neibs(node):
if (
type(node.op) is Images2Neibs
and node.inputs[0].dtype == "float32"
and node.op.mode in ["valid", "ignore_borders", "wrap_centered"]
):
return [
host_from_gpu(
gpu_images2neibs(gpu_from_host(node.inputs[0]), node.inputs[1], node.inputs[2], mode=node.op.mode)
)
]示例10: save_weights
def save_weights(weights, filename):
""" Taken from the convnet code. Deals with network calculated
on a gpu
"""
length = reduce(lambda x, y: x * y, weights.shape.eval())
data = host_from_gpu(weights).eval()
data = np.asarray(data)
data = data.reshape(length)
data = "\n".join([str(i) for i in data])
f = open(filename, "w")
f.write(data)
f.close()示例11: save_data
def save_data(self, filename, data, gpu = False):
mult = lambda x, y: x * y
if gpu:
length = reduce(mult, data.shape.eval())
data = host_from_gpu(data).eval()
data = np.asarray(data)
else:
length = reduce(mult, data.shape)
data = data.reshape(length)
data = "\n".join([str(i) for i in data])
f = open(filename, "w")
f.write(data)
f.close()示例12: grab_cpu_scalar
def grab_cpu_scalar(v, nd):
if v.owner is not None:
n = v.owner
if (isinstance(n.op, GpuDimShuffle) and
n.op.new_order == ('x',) * nd):
return host_from_gpu(n.inputs[0])
elif (isinstance(n.op, DimShuffle) and
n.op.new_order == ('x',) * nd):
return n.inputs[0]
elif isinstance(n.op, GpuFromHost):
return grab_cpu_scalar(n.inputs[0], nd=nd)
else:
return None
else:
if (isinstance(v, Constant) and
v.broadcastable == (True,) * nd):
return v.dimshuffle(())示例13: local_gpu_forloop
def local_gpu_forloop(node):
if isinstance(node.op, forloop):
sw = False
for inp in node.inputs:
if inp.owner and inp.owner.op == host_from_gpu:
sw = True
if sw:
inps = node.inputs
nw_inps = []
for inp in inps:
if not isinstance(inp.type, CudaNdarrayType):
nw_inps.append(gpu_from_host(inp))
else:
nw_inps.append(inp)
new_outs = node.op(*nw_inps)
return [host_from_gpu(x) for x in new_outs]
else:
return False示例14: local_gpu_multinomial
def local_gpu_multinomial(node):
if type(node.op) is MultinomialFromUniform:
p, u = node.inputs
m, = node.outputs
if (p.dtype == u.dtype == m.dtype == 'float32' and
any([i.owner and isinstance(i.owner.op, theano.sandbox.cuda.HostFromGpu)
for i in node.inputs])):
gpu_op = GpuMultinomialFromUniform(node.op.odtype)
return [host_from_gpu(gpu_op(*[gpu_from_host(i) for i in node.inputs])).T]
if (isinstance(node.op, theano.sandbox.cuda.GpuFromHost) and
node.inputs[0].owner and type(node.inputs[0].owner.op) is MultinomialFromUniform):
multi = node.inputs[0].owner
p, u = multi.inputs
m, = multi.outputs
if (p.dtype == u.dtype == m.dtype == 'float32'):
gpu_op = GpuMultinomialFromUniform(multi.op.odtype)
ret = gpu_op(*[gpu_from_host(i) for i in multi.inputs]).T
# The dimshuffle is on the cpu, but will be moved to the gpu by an opt.
return [gpu_from_host(ret)]示例15: local_assigner
def local_assigner(node):
if type(node.op) is Assigner:
p, indx, gr, = node.inputs
vals, = node.outputs
if (p.dtype == vals.dtype == 'float32' and
any([i.owner and isinstance(i.owner.op, theano.sandbox.cuda.HostFromGpu) for i in node.inputs])):
gpu_op = GpuAssigner()
ret = gpu_op(gpu_from_host(p),indx,gpu_from_host(gr))
return [host_from_gpu(ret),]
if (isinstance(node.op, theano.sandbox.cuda.GpuFromHost) and
node.inputs[0].owner and type(node.inputs[0].owner.op)
is Assigner):
multi = node.inputs[0].owner
p,indx,gr = multi.inputs
vals, = multi.outputs
if (p.dtype == vals.dtype == 'float32'):
gpu_op = GpuAssigner()
ret_vals = gpu_op(gpu_from_host(p),indx,gpu_from_host(gr))
return [gpu_from_host(ret_vals)]