提交 a39948d3 authored 作者: Frederic's avatar Frederic

Better doc of BatchedDotOp

上级 ee7fa8b5
...@@ -19,6 +19,7 @@ Blas Op ...@@ -19,6 +19,7 @@ Blas Op
.. automodule:: theano.sandbox.cuda.blas .. automodule:: theano.sandbox.cuda.blas
:members: :members:
.. autofunction:: theano.sandbox.cuda.blas.batched_dot
Nnet Op Nnet Op
======= =======
......
...@@ -16,10 +16,6 @@ from theano.tensor import as_tensor_variable ...@@ -16,10 +16,6 @@ from theano.tensor import as_tensor_variable
class BatchedDotOp(GpuOp): class BatchedDotOp(GpuOp):
"""
Call cublasSgemmBatched. Take 2 3d tensor as input.
"""
__props__ = () __props__ = ()
def make_node(self, inp1, inp2): def make_node(self, inp1, inp2):
...@@ -217,6 +213,10 @@ class BatchedDotOp(GpuOp): ...@@ -217,6 +213,10 @@ class BatchedDotOp(GpuOp):
return (1,) return (1,)
batched_dot = BatchedDotOp() batched_dot = BatchedDotOp()
"""
Call cublasSgemmBatched. Take 2 3d tensor as input.
"""
class GpuDot22(GpuOp): class GpuDot22(GpuOp):
""" """
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论