def type_inference(self):
if any_symbolic(self.shape.shape):
# We can't infer any shape if shape has variable length.
return types.tensor(self.x.dtype, (get_new_variadic_symbol(),))
# shape has fixed length here.
if self.shape.sym_val is None:
shape = tuple([get_new_symbol() for _ in range(self.shape.shape[0])])
return types.tensor(self.x.dtype, shape)
t, _ = self._get_type_val()
return t
def type_inference(self):
if any_symbolic(self.shape.shape):
# We can't infer any shape if shape has variable length.
return types.tensor(types.fp32, (get_new_variadic_symbol(),))
# shape has fixed length here.
if self.shape.sym_val is None:
ret_shape = tuple([get_new_symbol() for _ in range(self.shape.shape[0])])
return types.tensor(types.fp32, ret_shape)
return types.tensor(self.value.dtype, tuple(self.shape.sym_val.tolist()))
def test_builder_to_backend_symbolic(self, use_cpu_only, backend):
s0 = get_new_symbol()
s_len = get_new_symbol()
s1 = get_new_variadic_symbol()
input_placeholders = {
"x": mb.placeholder(shape=(2, s0)),
"shape": mb.placeholder(shape=(3, ), dtype=types.int32),
"shape2": mb.placeholder(shape=(s_len, ), dtype=types.int32),
}
def build(x, shape, shape2):
return [
mb.reshape(x=x, shape=[2, -1]),
mb.reshape(x=x, shape=[1, -1]),
mb.reshape(x=x, shape=[2, 1, 1, -1]),
mb.reshape(x=x, shape=shape),
mb.reshape(x=x, shape=shape2),
]
expected_output_types = [
(2, s0, types.fp32),
(1, 2 * s0, types.fp32),
(2, 1, 1, s0, types.fp32),
(UNK_SYM, UNK_SYM, UNK_SYM, types.fp32),
(UNK_VARIADIC, types.fp32),
]
expected_outputs = [
np.array([[1, 2, 3], [4, 5, 6]], dtype=np.float32),
np.array([[1, 2, 3, 4, 5, 6]], dtype=np.float32),
np.array([[[[1.0, 2.0, 3.0]]], [[[4.0, 5.0, 6.0]]]],
dtype=np.float32),
np.array([[[1, 2, 3]], [[4, 5, 6]]], dtype=np.float32),
np.array([[[1, 2, 3]], [[4, 5, 6]]], dtype=np.float32),
]
input_values = {
"x": np.array([[1, 2, 3], [4, 5, 6]], dtype=np.float32),
"shape": np.array([2, 1, 3], dtype=np.float32),
"shape2": np.array([2, 1, 3], dtype=np.float32),
}
run_compare_builder(
build,
input_placeholders,
input_values,
expected_output_types,
expected_outputs,
use_cpu_only=use_cpu_only,
frontend_only=False,
backend=backend,
)
def type_inference(self):
x_type = self.x.dtype
perm = self.perm.val
x_shape = np.array(self.x.shape)
if len(perm) != self.x.rank:
msg = "perm should have the same length as rank(x): {} != {}"
raise ValueError(msg.format(len(perm), self.x.rank))
if self.x.rank == 0:
return self.x.sym_type # scalar cannot be transposed
if any_variadic(self.x.shape):
ret_shape = get_new_variadic_symbol()
else:
ret_shape = x_shape[perm]
return types.tensor(x_type, tuple(ret_shape))
