def getTestCases(self):
return {
"pad_4d_2d": [
base.Tensor([1, 2, 2, 1]),
base.Tensor([4, 2], dtype=tf.int32, const_val=[0, 0, 1, 1, 1, 1, 0, 0])
]
}
def getTestCases(self):
'''
this returns a a hash containg test cases.
key of return hash is test case name and
value of return hash is test is a list of input tensor metadata.
test name (key of hash) is used as
- prefix of file name to be generated (don't use white space or special characters)
- output node name pf graph
'''
return {"stack_4d": [base.Tensor([1, 4, 3]), base.Tensor([1, 4, 3])]}
def getTestCases(self):
'''
this returns a hash of test case (= set of input type), for example:
[1.2, -2.3] : two input, both are scalar. one is 1.2, another is -2.3
[[5,3], [5,4,3]] : two input, both are shapes. one is [5.3], another is [5,4,3]
test name (key of hash) is used as
- prefix of file name to be generated
- output node name pf graph
'''
return {
"topk_2d": [
base.Tensor(shape=[2, 3], dtype=tf.float32),
base.Tensor(shape=[], const_val=2, dtype=tf.int32)
],
"topk_3d": [
base.Tensor(shape=[2, 3, 4], dtype=tf.float32),
base.Tensor(shape=[], const_val=2, dtype=tf.int32)
],
}
def getTestCases(self):
'''
this returns a a hash containg test cases.
key of return hash is test case name and
value of return hash is test is a list of input tensor metadata.
test name (key of hash) is used as
- prefix of file name to be generated (don't use white space or special characters)
- output node name pf graph
'''
# yapf: disable
return {
"div_scalarConst_scalarConst":
[base.Tensor([], const_val=1.2),
base.Tensor([], const_val=-2.3)],
"div_1d_1d": [base.Tensor([5]), base.Tensor([5])],
"div_2d_2d": [base.Tensor([5, 3]), base.Tensor([5, 3])],
"div_3d_3d": [base.Tensor([5, 4, 3]),
base.Tensor([5, 4, 3])],
"div_4d_4d": [base.Tensor([2, 5, 4, 3]),
base.Tensor([2, 5, 4, 3])],
# broadcasting by scalar
"div_1d_scalarConst": [base.Tensor([5]),
base.Tensor([], const_val=1.1)],
"div_2d_scalarConst": [base.Tensor([5, 3]),
base.Tensor([], const_val=1.1)],
"div_3d_scalarConst": [base.Tensor([5, 4, 3]),
base.Tensor([], const_val=1.1)],
"div_4d_scalarConst": [base.Tensor([2, 5, 4, 3]),
base.Tensor([], const_val=1.1)],
# broadcasting by 1d
"div_2d_1d": [base.Tensor([5, 3]),
base.Tensor( [3])],
"div_3d_1d": [base.Tensor([5, 4, 3]),
base.Tensor( [3])],
"div_4d_1d": [base.Tensor([2, 5, 4, 3]),
base.Tensor( [3])],
# broadcasting by 2d
"div_3d_2d": [base.Tensor([5, 4, 3]),
base.Tensor( [4, 3])],
"div_4d_2d": [base.Tensor([2, 5, 4, 3]),
base.Tensor( [4, 3])],
# broadcasting by 3d
"div_4d_3d": [base.Tensor([2, 5, 4, 3]),
base.Tensor( [5, 4, 3])]
}
def getTestCases(self):
return {"floor_4d_4d": [base.Tensor([1, 2, 2, 1]), base.Tensor([1, 2, 2, 1])]}
def getTestCases(self):
return {"transpose_4d": [base.Tensor([1, 2, 2, 1])]}
def getTestCases(self):
return {"squeeze_3d": [base.Tensor([1, 5, 1])]}