Skip to content

P
pytensor
提交 98d2a0c3 authored 作者: Frederic Bastien's avatar Frederic Bastien
浏览文件
操作

In Image2Neibs(images2neibs) op added the neib_step parameter and updated the…

In Image2Neibs(images2neibs) op added the neib_step parameter and updated the gpu op to raise an error if it is not the default as it don't implement the new option.
上级 bb466e6d
显示空白字符变更
内嵌 并排
正在显示 包含 169 行增加 和 20 行删除
theano/sandbox/neighbours.py
浏览文件 @ 98d2a0c3
...@@ -17,18 +17,32 @@ class Images2Neibs(Op): ...@@ -17,18 +17,32 @@ class Images2Neibs(Op):
return hash(type(self)) return hash(type(self))
def __str__(self): def __str__(self):
return self.__class__.__name__ return self.__class__.__name__
def make_node(self, ten4, neib_shape): def make_node(self, ten4, neib_shape, neib_step=None):
"""
:param neib_step: (dx,dy) where dx is the number of rows to skip between patch
and dy is the number of columns. When None, this is the same
as neib_shape(patch are disjoint)
"""
ten4 = T.as_tensor_variable(ten4) ten4 = T.as_tensor_variable(ten4)
neib_shape = T.as_tensor_variable(neib_shape) neib_shape = T.as_tensor_variable(neib_shape)
return Apply(self, [ten4, neib_shape], [T.matrix(dtype=ten4.type.dtype)]) if neib_step is None:
neib_step = neib_shape
else:
neib_step = T.as_tensor_variable(neib_step)
assert ten4.ndim==4
assert neib_shape.ndim==1
assert neib_step.ndim==1
return Apply(self, [ten4, neib_shape,neib_step], [T.matrix(dtype=ten4.type.dtype)])
def grad(self, (pvals, unis), (gz,)): def grad(self, (pvals, unis), (gz,)):
return [None, None] return [None, None]
def c_code_cache_version(self): def c_code_cache_version(self):
return (2,) return (3,)
def c_code(self, node, name, (ten4, neib_shape), (z,), sub): def c_code(self, node, name, (ten4, neib_shape, neib_step), (z,), sub):
fail = sub['fail'] fail = sub['fail']
return """ return """
...@@ -48,25 +62,44 @@ class Images2Neibs(Op): ...@@ -48,25 +62,44 @@ class Images2Neibs(Op):
PyErr_Format(PyExc_TypeError, "neib_shape wrong shape ; has to contain 2 elements"); PyErr_Format(PyExc_TypeError, "neib_shape wrong shape ; has to contain 2 elements");
%(fail)s; %(fail)s;
} }
if (%(neib_step)s->nd != 1)
{
PyErr_Format(PyExc_TypeError, "neib_step wrong rank");
%(fail)s;
}
if ( (%(neib_step)s->dimensions)[0] != 2)
{
PyErr_Format(PyExc_TypeError, "neib_step wrong step ; has to contain 2 elements");
%(fail)s;
}
// (c,d) = neib_shape
const npy_intp c = (npy_intp) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 0); const npy_intp c = (npy_intp) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 0);
const npy_intp d = (npy_intp) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 1); const npy_intp d = (npy_intp) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 1);
// (step_x,step_y) = neib_step
const npy_intp step_x = (npy_intp) *(dtype_%(neib_step)s*) PyArray_GETPTR1(%(neib_step)s, 0);
const npy_intp step_y = (npy_intp) *(dtype_%(neib_step)s*) PyArray_GETPTR1(%(neib_step)s, 1);
if ( (%(ten4)s->dimensions)[2] %% c != 0) if ( ((%(ten4)s->dimensions)[2] < c) ||( (((%(ten4)s->dimensions)[2]-c) %% step_x)!=0))
{ {
PyErr_Format(PyExc_TypeError, "neib_shape[0] must divide ten4.shape[2]"); PyErr_Format(PyExc_TypeError, "neib_shape[0]=%%d, neib_step[0]=%%d and ten4.shape[2]=%%d not consistent",
c, step_x, %(ten4)s->dimensions[2]);
%(fail)s; %(fail)s;
} }
if ( (%(ten4)s->dimensions)[3] %% d != 0) if ( ((%(ten4)s->dimensions)[3] < d) ||( (((%(ten4)s->dimensions)[3]-d) %% step_y)!=0))
{ {
PyErr_Format(PyExc_TypeError, "neib_shape[1] must divide ten4.shape[3]"); PyErr_Format(PyExc_TypeError, "neib_shape[1]=%%d, neib_step[1]=%%d and ten4.shape[3]=%%d not consistent",
d, step_y, %(ten4)s->dimensions[3]);
%(fail)s; %(fail)s;
} }
const int grid_c = 1+(((%(ten4)s->dimensions)[2]-c)/step_x); //number of patch in height
const int grid_d = 1+(((%(ten4)s->dimensions)[3]-d)/step_y); //number of patch in width
// new dimensions for z // new dimensions for z
const npy_intp z_dim1 = c * d; const npy_intp z_dim1 = c * d;
const npy_intp z_dim0 = (%(ten4)s->dimensions)[2] / c const npy_intp z_dim0 = grid_c
* (%(ten4)s->dimensions)[3] / d * grid_d
* (%(ten4)s->dimensions)[1] * (%(ten4)s->dimensions)[1]
* (%(ten4)s->dimensions)[0]; * (%(ten4)s->dimensions)[0];
...@@ -101,26 +134,29 @@ class Images2Neibs(Op): ...@@ -101,26 +134,29 @@ class Images2Neibs(Op):
const int width = (%(ten4)s->dimensions)[3]; const int width = (%(ten4)s->dimensions)[3];
// (c,d) = neib_shape // (c,d) = neib_shape
const int c = (int) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 0); const npy_intp c = (npy_intp) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 0);
const int d = (int) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 1); const npy_intp d = (npy_intp) *(dtype_%(neib_shape)s*) PyArray_GETPTR1(%(neib_shape)s, 1);
// (step_x,step_y) = neib_step
const npy_intp step_x = (npy_intp) *(dtype_%(neib_step)s*) PyArray_GETPTR1(%(neib_step)s, 0);
const npy_intp step_y = (npy_intp) *(dtype_%(neib_step)s*) PyArray_GETPTR1(%(neib_step)s, 1);
const int grid_c = 1+(((%(ten4)s->dimensions)[2]-c)/step_x); //number of patch in height
const int grid_d = 1+(((%(ten4)s->dimensions)[3]-d)/step_y); //number of patch in width
const int grid_c = height/c;
const int grid_d = width/d;
// Oh this is messed up... // Oh this is messed up...
for (int n = 0; n < nb_batch; n++) // loop over batches for (int n = 0; n < nb_batch; n++) // loop over batches
for (int s = 0; s < nb_stack; s++) // loop over stacks for (int s = 0; s < nb_stack; s++) // loop over stacks
for (int a = 0; a < grid_c; a++) // loop over height/c for (int a = 0; a < grid_c; a++) // loop over the number of patch in height
for (int b = 0; b < grid_d; b++) // loop over width/d for (int b = 0; b < grid_d; b++) // loop over the number of patch in width
{ {
int z_row = b + grid_d*(a + grid_c*(s + nb_stack*n)); int z_row = b + grid_d*(a + grid_c*(s + nb_stack*n));
for (int i = 0; i < c; i++) // loop over c for (int i = 0; i < c; i++) // loop over c
{ {
int ten4_2 = i + a * c; int ten4_2 = i + a * step_x;
for (int j = 0; j < d; j++) // loop over d for (int j = 0; j < d; j++) // loop over d
{ {
int ten4_3 = j + b * d; int ten4_3 = j + b * step_y;
int z_col = j + d * i; int z_col = j + d * i;
dtype_%(z)s* curr_z = (dtype_%(z)s*) PyArray_GETPTR2(%(z)s, z_row, z_col); dtype_%(z)s* curr_z = (dtype_%(z)s*) PyArray_GETPTR2(%(z)s, z_row, z_col);
...@@ -156,7 +192,9 @@ def neibs2images(neibs, neib_shape, original_shape): ...@@ -156,7 +192,9 @@ def neibs2images(neibs, neib_shape, original_shape):
# This is work in progress # This is work in progress
class GpuImages2Neibs(Images2Neibs): class GpuImages2Neibs(Images2Neibs):
def make_node(self, ten4, neib_shape): def make_node(self, ten4, neib_shape, neib_step):
if neib_shape!=neib_step:
raise NotImplementedError("neib_step not implemented now on the gpu")
assert ten4.dtype == 'float32' assert ten4.dtype == 'float32'
#assert neib_shape.dtype == 'float32' #assert neib_shape.dtype == 'float32'
if not isinstance(ten4.type, CudaNdarrayType): if not isinstance(ten4.type, CudaNdarrayType):
...@@ -169,6 +207,7 @@ class GpuImages2Neibs(Images2Neibs): ...@@ -169,6 +207,7 @@ class GpuImages2Neibs(Images2Neibs):
dtype=ten4.type.dtype)()]) dtype=ten4.type.dtype)()])
def c_code_cache_version(self): def c_code_cache_version(self):
return ()
return (2,) return (2,)
def c_support_code_apply(self, node, nodename): def c_support_code_apply(self, node, nodename):
...@@ -368,7 +407,7 @@ gpu_images2neibs = GpuImages2Neibs() ...@@ -368,7 +407,7 @@ gpu_images2neibs = GpuImages2Neibs()
@local_optimizer() @local_optimizer()
def use_gpu_images2neibs(node): def use_gpu_images2neibs(node):
if node.op == images2neibs: if node.op == images2neibs:
return [host_from_gpu(gpu_images2neibs(*[gpu_from_host(node.inputs[0]),node.inputs[1]]))] return [host_from_gpu(gpu_images2neibs(*[gpu_from_host(node.inputs[0]),node.inputs[1],node.inputs[2]]))]
if cuda_available: if cuda_available:
register_gpu_opt()(use_gpu_images2neibs) register_gpu_opt()(use_gpu_images2neibs)
......
theano/sandbox/test_neighbours.py
浏览文件 @ 98d2a0c3
...@@ -29,6 +29,116 @@ def test_neibs(): ...@@ -29,6 +29,116 @@ def test_neibs():
#print g() #print g()
assert numpy.allclose(images.value,g()) assert numpy.allclose(images.value,g())
def test_neibs_bad_shape():
shape = (2,3,10,10)
images = shared(numpy.arange(numpy.prod(shape)).reshape(shape))
neib_shape = T.as_tensor_variable((3,2))
try:
f = function([], images2neibs(images, neib_shape), mode=mode_without_gpu)
neibs = f()
#print neibs
assert False,"An error was expected"
except TypeError:
pass
shape = (2,3,10,10)
images = shared(numpy.arange(numpy.prod(shape)).reshape(shape))
neib_shape = T.as_tensor_variable((2,3))
try:
f = function([], images2neibs(images, neib_shape), mode=mode_without_gpu)
neibs = f()
#print neibs
assert False,"An error was expected"
except TypeError:
pass
def test_neibs_manual():
shape = (2,3,4,4)
images = shared(numpy.arange(numpy.prod(shape)).reshape(shape))
neib_shape = T.as_tensor_variable((2,2))
f = function([], images2neibs(images, neib_shape), mode=mode_without_gpu)
#print images.value
neibs = f()
print neibs
assert numpy.allclose(neibs,[[ 0, 1, 4, 5],
[ 2, 3, 6, 7],
[ 8, 9, 12, 13],
[10, 11, 14, 15],
[16, 17, 20, 21],
[18, 19, 22, 23],
[24, 25, 28, 29],
[26, 27, 30, 31],
[32, 33, 36, 37],
[34, 35, 38, 39],
[40, 41, 44, 45],
[42, 43, 46, 47],
[48, 49, 52, 53],
[50, 51, 54, 55],
[56, 57, 60, 61],
[58, 59, 62, 63],
[64, 65, 68, 69],
[66, 67, 70, 71],
[72, 73, 76, 77],
[74, 75, 78, 79],
[80, 81, 84, 85],
[82, 83, 86, 87],
[88, 89, 92, 93],
[90, 91, 94, 95]])
g = function([], neibs2images(neibs, neib_shape, images.shape), mode=mode_without_gpu)
print g()
assert numpy.allclose(images.value,g())
def test_neibs_manual_step():
shape = (2,3,5,5)
images = shared(numpy.asarray(numpy.arange(numpy.prod(shape)).reshape(shape),dtype='float32'))
neib_shape = T.as_tensor_variable((3,3))
neib_step = T.as_tensor_variable((2,2))
modes = [mode_without_gpu]
if cuda.cuda_available:
modes.append(mode_with_gpu)
for mode in modes:
f = function([], images2neibs(images, neib_shape, neib_step), mode=mode)
#print images.value
neibs = f()
print neibs
assert numpy.allclose(neibs,
[[ 0, 1, 2, 5, 6, 7, 10, 11, 12],
[ 2, 3, 4, 7, 8, 9, 12, 13, 14],
[ 10, 11, 12, 15, 16, 17, 20, 21, 22],
[ 12, 13, 14, 17, 18, 19, 22, 23, 24],
[ 25, 26, 27, 30, 31, 32, 35, 36, 37],
[ 27, 28, 29, 32, 33, 34, 37, 38, 39],
[ 35, 36, 37, 40, 41, 42, 45, 46, 47],
[ 37, 38, 39, 42, 43, 44, 47, 48, 49],
[ 50, 51, 52, 55, 56, 57, 60, 61, 62],
[ 52, 53, 54, 57, 58, 59, 62, 63, 64],
[ 60, 61, 62, 65, 66, 67, 70, 71, 72],
[ 62, 63, 64, 67, 68, 69, 72, 73, 74],
[ 75, 76, 77, 80, 81, 82, 85, 86, 87],
[ 77, 78, 79, 82, 83, 84, 87, 88, 89],
[ 85, 86, 87, 90, 91, 92, 95, 96, 97],
[ 87, 88, 89, 92, 93, 94, 97, 98, 99],
[100, 101, 102, 105, 106, 107, 110, 111, 112],
[102, 103, 104, 107, 108, 109, 112, 113, 114],
[110, 111, 112, 115, 116, 117, 120, 121, 122],
[112, 113, 114, 117, 118, 119, 122, 123, 124],
[125, 126, 127, 130, 131, 132, 135, 136, 137],
[127, 128, 129, 132, 133, 134, 137, 138, 139],
[135, 136, 137, 140, 141, 142, 145, 146, 147],
[137, 138, 139, 142, 143, 144, 147, 148, 149]])
#g = function([], neibs2images(neibs, neib_shape, images.shape), mode=mode_without_gpu)
#print g()
#assert numpy.allclose(images.value,g())
def test_neibs_gpu(): def test_neibs_gpu():
if cuda.cuda_available == False: if cuda.cuda_available == False:
raise SkipTest('Optional package cuda disabled') raise SkipTest('Optional package cuda disabled')
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论