def __getitem_(g, self, i):
if sym_help._is_tensor_list(self):
# SequenceAt requires that the input be a List of Tensors
return g.op("SequenceAt", self, i)
else:
from torch.onnx.symbolic_opset9 import __getitem_ as getitem
return getitem(g, self, i)
def _len(g, self):
if (
symbolic_helper._is_tensor_list(self)
or self.node().kind() == "onnx::SplitToSequence"
):
return g.op("SequenceLength", self)
sz_0 = size(g, self, g.op("Constant", value_t=torch.LongTensor([0])))
return symbolic_helper._squeeze_helper(g, sz_0, [0])
def add(g, self, other, alpha=None):
if sym_help._is_value(self) and sym_help._is_tensor_list(self):
tensor_list_node = other.node()
if tensor_list_node.kind() != "prim::ListConstruct":
return _unimplemented("add", "does not support adding dynamic tensor list to another")
tensors = sym_help._unpack_list(other)
l = self
for t in tensors:
l = g.op("SequenceInsert", l, t)
return l
return torch.onnx.symbolic_opset9.add(g, self, other, alpha)
def verify_inferred_shape(graph):
# Check every node in graph has type properly assigned.
for n in graph.nodes():
for out in n.outputs():
if not _is_tensor_list(out) and not _is_tensor(
out) and not _is_none(out):
raise RuntimeError(
"Output of node is neither type Tensor nor type list of Tensor: ",
out)
if _is_tensor(out) and out.type().scalarType() is None:
raise RuntimeError(
"Output of node does not have type assigned", out)
if _is_tensor(out) and out.type().dim() is None:
raise RuntimeError(
"Output of node does not have shape assigned", out)
def _len(g, self):
if _is_tensor_list(self) or self.node().kind() == "onnx::SplitToSequence":
return g.op("SequenceLength", self)
return g.op("Size", self)
def _len(g, self):
if _is_tensor_list(self) or self.node().kind() == "onnx::SplitToSequence":
return g.op("SequenceLength", self)
sz_0 = size(g, self, g.op("Constant", value_t=torch.LongTensor([0])))
return g.op('Squeeze', sz_0, axes_i=[0])
