Skip to content
项目
群组
代码片段
帮助
当前项目
正在载入...
登录 / 注册
切换导航面板
P
pytensor
项目
项目
详情
活动
周期分析
仓库
仓库
文件
提交
分支
标签
贡献者
图表
比较
统计图
议题
0
议题
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
CI / CD
CI / CD
流水线
作业
日程
统计图
Wiki
Wiki
代码片段
代码片段
成员
成员
折叠边栏
关闭边栏
活动
图像
聊天
创建新问题
作业
提交
问题看板
Open sidebar
testgroup
pytensor
Commits
8e633dca
提交
8e633dca
authored
4月 25, 2012
作者:
Yann N. Dauphin
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
allow for duplicates and unsorted sparse dimensions
上级
76c598ee
隐藏空白字符变更
内嵌
并排
正在显示
1 个修改的文件
包含
52 行增加
和
55 行删除
+52
-55
basic.py
theano/sparse/basic.py
+52
-55
没有找到文件。
theano/sparse/basic.py
浏览文件 @
8e633dca
...
@@ -698,12 +698,12 @@ class CSM(gof.Op):
...
@@ -698,12 +698,12 @@ class CSM(gof.Op):
indptr
.
copy
()),
shape
.
copy
(),
indptr
.
copy
()),
shape
.
copy
(),
copy
=
False
)
copy
=
False
)
def
grad
(
self
,
(
x_data
,
x_indices
,
x_indptr
,
_
),
(
g_out
,)):
def
grad
(
self
,
(
x_data
,
x_indices
,
x_indptr
,
x_shape
),
(
g_out
,)):
"""Return a gradient on the data vector"""
"""Return a gradient on the data vector"""
g_data
,
g_indices
,
g_indptr
,
_
=
csm_properties
(
g_out
)
g_data
,
g_indices
,
g_indptr
,
g_shape
=
csm_properties
(
g_out
)
#unpack the data vector and wrap it as a 1d TensorType
#unpack the data vector and wrap it as a 1d TensorType
g_data
=
csm_grad
(
self
.
kmap
)(
x_data
,
x_indices
,
x_indptr
,
g_data
=
csm_grad
(
self
.
kmap
)(
x_data
,
x_indices
,
x_indptr
,
x_shape
,
g_data
,
g_indices
,
g_indptr
)
g_data
,
g_indices
,
g_indptr
,
g_shape
)
return
[
g_data
,
None
,
None
,
None
]
return
[
g_data
,
None
,
None
,
None
]
def
infer_shape
(
self
,
node
,
shapes
):
def
infer_shape
(
self
,
node
,
shapes
):
...
@@ -735,33 +735,30 @@ class CSMGrad(gof.op.Op):
...
@@ -735,33 +735,30 @@ class CSMGrad(gof.op.Op):
self
.
__class__
.
__name__
,
self
.
__class__
.
__name__
,
self
.
kmap
)
self
.
kmap
)
def
make_node
(
self
,
x_data
,
x_indices
,
x_indptr
,
def
make_node
(
self
,
x_data
,
x_indices
,
x_indptr
,
x_shape
,
g_data
,
g_indices
,
g_indptr
):
g_data
,
g_indices
,
g_indptr
,
g_shape
):
gout_data
=
g_data
.
type
()
gout_data
=
g_data
.
type
()
return
gof
.
Apply
(
self
,
[
x_data
,
x_indices
,
x_indptr
,
return
gof
.
Apply
(
self
,
[
x_data
,
x_indices
,
x_indptr
,
x_shape
,
g_data
,
g_indices
,
g_indptr
],
[
gout_data
])
g_data
,
g_indices
,
g_indptr
,
g_shape
],
[
gout_data
])
def
perform
(
self
,
node
,
(
x_data
,
x_indices
,
x_indptr
,
def
perform
(
self
,
node
,
(
x_data
,
x_indices
,
x_indptr
,
x_shape
,
g_data
,
g_indices
,
g_indptr
),
(
g_out
,)):
g_data
,
g_indices
,
g_indptr
,
g_shape
),
(
g_out
,)):
if
len
(
x_indptr
)
-
1
==
x_shape
[
0
]:
sp_dim
=
x_shape
[
1
]
else
:
sp_dim
=
x_shape
[
0
]
g_row
=
numpy
.
zeros
(
sp_dim
,
dtype
=
g_data
.
dtype
)
gout_data
=
numpy
.
zeros_like
(
x_data
)
gout_data
=
numpy
.
zeros_like
(
x_data
)
for
i
in
range
(
len
(
x_indptr
)
-
1
):
for
i
in
range
(
len
(
x_indptr
)
-
1
):
x_pos
=
x_indptr
[
i
]
for
j_ptr
in
range
(
g_indptr
[
i
],
g_indptr
[
i
+
1
]):
g_pos
=
g_indptr
[
i
]
g_row
[
g_indices
[
j_ptr
]]
+=
g_data
[
j_ptr
]
x_end
=
x_indptr
[
i
+
1
]
g_end
=
g_indptr
[
i
+
1
]
while
x_pos
<
x_end
and
g_pos
<
g_end
:
for
j_ptr
in
range
(
x_indptr
[
i
],
x_indptr
[
i
+
1
]):
x_ind
=
x_indices
[
x_pos
]
gout_data
[
j_ptr
]
=
g_row
[
x_indices
[
j_ptr
]]
g_ind
=
g_indices
[
g_pos
]
for
j_ptr
in
range
(
g_indptr
[
i
],
g_indptr
[
i
+
1
]):
if
x_ind
==
g_ind
:
g_row
[
g_indices
[
j_ptr
]]
=
0
gout_data
[
x_pos
]
=
g_data
[
g_pos
]
x_pos
+=
1
g_pos
+=
1
elif
x_ind
<
g_ind
:
x_pos
+=
1
else
:
g_pos
+=
1
if
self
.
kmap
is
None
:
if
self
.
kmap
is
None
:
g_out
[
0
]
=
gout_data
g_out
[
0
]
=
gout_data
...
@@ -789,19 +786,19 @@ class CSMGradC(gof.Op):
...
@@ -789,19 +786,19 @@ class CSMGradC(gof.Op):
def
__str__
(
self
):
def
__str__
(
self
):
return
self
.
__class__
.
__name__
return
self
.
__class__
.
__name__
def
make_node
(
self
,
a_val
,
a_ind
,
a_ptr
,
b_val
,
b_ind
,
b_ptr
):
def
make_node
(
self
,
a_val
,
a_ind
,
a_ptr
,
a_dim
,
b_val
,
b_ind
,
b_ptr
,
b_dim
):
return
gof
.
Apply
(
self
,
[
a_val
,
a_ind
,
a_ptr
,
b_val
,
b_ind
,
b_ptr
]
,
return
gof
.
Apply
(
self
,
[
a_val
,
a_ind
,
a_ptr
,
a_dim
,
[
b_val
.
type
()])
b_val
,
b_ind
,
b_ptr
,
b_dim
],
[
b_val
.
type
()])
def
c_code
(
self
,
node
,
name
,
(
a_val
,
a_ind
,
a_ptr
,
def
c_code
(
self
,
node
,
name
,
(
a_val
,
a_ind
,
a_ptr
,
a_dim
,
b_val
,
b_ind
,
b_ptr
),
(
z
,),
sub
):
b_val
,
b_ind
,
b_ptr
,
b_dim
),
(
z
,),
sub
):
# retrieve dtype number
# retrieve dtype number
typenum_z
=
node
.
outputs
[
0
]
.
type
.
dtype_specs
()[
-
1
]
typenum_z
=
node
.
outputs
[
0
]
.
type
.
dtype_specs
()[
-
1
]
if
node
.
inputs
[
0
]
.
type
.
dtype
in
(
'complex64'
,
'complex128'
):
if
node
.
inputs
[
0
]
.
type
.
dtype
in
(
'complex64'
,
'complex128'
):
raise
NotImplementedError
(
'Complex types are not supported for a_val'
)
raise
NotImplementedError
(
'Complex types are not supported for a_val'
)
if
node
.
inputs
[
3
]
.
type
.
dtype
in
(
'complex64'
,
'complex128'
):
if
node
.
inputs
[
3
]
.
type
.
dtype
in
(
'complex64'
,
'complex128'
):
raise
NotImplementedError
(
'Complex types are not supported for b_val'
)
raise
NotImplementedError
(
'Complex types are not supported for b_val'
)
return
"""
return
"""
if (
%(a_val)
s->nd != 1) {PyErr_SetString(PyExc_NotImplementedError, "rank(a_val) != 1");
%(fail)
s;}
if (
%(a_val)
s->nd != 1) {PyErr_SetString(PyExc_NotImplementedError, "rank(a_val) != 1");
%(fail)
s;}
if (
%(a_ind)
s->nd != 1) {PyErr_SetString(PyExc_NotImplementedError, "rank(a_ind) != 1");
%(fail)
s;}
if (
%(a_ind)
s->nd != 1) {PyErr_SetString(PyExc_NotImplementedError, "rank(a_ind) != 1");
%(fail)
s;}
...
@@ -842,7 +839,11 @@ class CSMGradC(gof.Op):
...
@@ -842,7 +839,11 @@ class CSMGradC(gof.Op):
{
{
// sparse array has size MxK, dense KxN, output MxN
// sparse array has size MxK, dense KxN, output MxN
npy_intp M =
%(a_ptr)
s->dimensions[0] - 1;
npy_intp M =
%(a_ptr)
s->dimensions[0] - 1;
npy_intp a_dim_0 = ((npy_int32 *)
%(a_dim)
s->data)[0];
npy_intp a_dim_1 = ((npy_int32 *)
%(a_dim)
s->data)[1];
npy_intp sp_dim = (M == a_dim_0)?a_dim_1:a_dim_0;
// strides tell you how many bytes to skip to go to next column/row entry
// strides tell you how many bytes to skip to go to next column/row entry
npy_intp Sz =
%(z)
s->strides[0] /
%(z)
s->descr->elsize;
npy_intp Sz =
%(z)
s->strides[0] /
%(z)
s->descr->elsize;
npy_intp Sa_val =
%(a_val)
s->strides[0] /
%(a_val)
s->descr->elsize;
npy_intp Sa_val =
%(a_val)
s->strides[0] /
%(a_val)
s->descr->elsize;
...
@@ -862,33 +863,29 @@ class CSMGradC(gof.Op):
...
@@ -862,33 +863,29 @@ class CSMGradC(gof.Op):
const npy_int32 * __restrict__ Db_ptr = (npy_int32*)
%(b_ptr)
s->data;
const npy_int32 * __restrict__ Db_ptr = (npy_int32*)
%(b_ptr)
s->data;
npy_intp nnz =
%(a_ind)
s->dimensions[0];
npy_intp nnz =
%(a_ind)
s->dimensions[0];
dtype_
%(b_val)
s b_row[sp_dim];
//clear the output array
//clear the output array
memset(Dz, 0, nnz*sizeof(dtype_
%(z)
s));
memset(Dz, 0, nnz*sizeof(dtype_
%(z)
s));
memset(b_row, 0, sp_dim*sizeof(dtype_
%(b_val)
s));
// loop over inner dimension
// loop over inner dimension
for (npy_int64 m = 0; m < M; ++m)
for (npy_int64 m = 0; m < M; ++m)
{
{
npy_int32 a_pos = Da_ptr[m * Sa
_ptr];
for (npy_int32 j_ptr = Db_ptr[m * Sb
_ptr];
npy_int32 b_pos = Db_ptr[m * Sb_ptr];
j_ptr < Db_ptr[(m + 1) * Sb_ptr]; j_ptr++) {
npy_int32 a_end = Da_ptr[(m + 1) * Sa_ptr
];
b_row[Db_ind[j_ptr * Sb_ind]] += Db_val[j_ptr*Sb_val
];
npy_int32 b_end = Db_ptr[(m + 1) * Sb_ptr];
}
while (a_pos < a_end && b_pos < b_end) {
for (npy_int32 j_ptr = Da_ptr[m * Sa_ptr];
npy_int32 a_ind = Da_ind[a_pos * Sa_ind];
j_ptr < Da_ptr[(m + 1) * Sa_ptr]; j_ptr++) {
npy_int32 b_ind = Db_ind[b_pos * Sb_ind];
Dz[j_ptr*Sz] = b_row[Da_ind[j_ptr * Sa_ind]];
}
if (a_ind == b_ind) {
Dz[a_pos*Sz] = Db_val[b_pos*Sb_val];
for (npy_int32 j_ptr = Db_ptr[m * Sb_ptr];
a_pos++;
j_ptr < Db_ptr[(m + 1) * Sb_ptr]; j_ptr++) {
b_pos++;
b_row[Db_ind[j_ptr * Sb_ind]] = 0;
}
else if (a_ind < b_ind) {
a_pos++;
}
else {
b_pos++;
}
}
}
}
}
}
}
...
@@ -896,12 +893,12 @@ class CSMGradC(gof.Op):
...
@@ -896,12 +893,12 @@ class CSMGradC(gof.Op):
"""
%
dict
(
locals
(),
**
sub
)
"""
%
dict
(
locals
(),
**
sub
)
def
c_code_cache_version
(
self
):
def
c_code_cache_version
(
self
):
return
(
1
,)
return
(
2
,)
csm_grad_c
=
CSMGradC
()
csm_grad_c
=
CSMGradC
()
@gof.local_optimizer
([
csm_grad
(
None
)])
@gof.local_optimizer
([
csm_grad
(
None
)])
def
local_csm_grad_c
(
node
):
def
local_csm_grad_c
(
node
):
"""
usmm -> usmm_csc_dense
"""
"""
csm_grad(None) -> csm_grad_c
"""
if
node
.
op
==
csm_grad
(
None
):
if
node
.
op
==
csm_grad
(
None
):
return
[
csm_grad_c
(
*
node
.
inputs
)]
return
[
csm_grad_c
(
*
node
.
inputs
)]
return
False
return
False
...
...
编写
预览
Markdown
格式
0%
重试
或
添加新文件
添加附件
取消
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!
取消
请
注册
或者
登录
后发表评论