From f5212f52ce081611bd78dfd224c2fd4dd4ed263b Mon Sep 17 00:00:00 2001
From: Octave Perrin <operrin@lrtechnologies.fr>
Date: Fri, 29 Nov 2024 11:26:27 +0100
Subject: [PATCH] more python bind
---
include/aidge/graph/GraphView.hpp | 6 +-
python_binding/graph/pybind_GraphView.cpp | 71 +++++++++++++++++++++--
2 files changed, 69 insertions(+), 8 deletions(-)
diff --git a/include/aidge/graph/GraphView.hpp b/include/aidge/graph/GraphView.hpp
index 921957e29..6d4a8cb28 100644
--- a/include/aidge/graph/GraphView.hpp
+++ b/include/aidge/graph/GraphView.hpp
@@ -296,9 +296,10 @@ public:
* compatible with the selected kernel.
* If not, add a Transpose Operator.
* 4 - Propagate Tensor dimensions through the consecutive Operators.
- @params string: backend Backend used, default is cpu
+ @params backend: Backend used, default is cpu
@params Aidge Datatype: datatype used, default is float32
- @params vector of vector of DimSize_t: dims
+ @params device: Device to be set
+ @params dims: vector of vector of DimSize_t: dims
*/
void compile(const std::string& backend = "cpu",
const Aidge::DataType datatype = DataType::Float32,
@@ -313,6 +314,7 @@ public:
* - Updates are made in node dependencies order, because if dims have changed
* at any point in the graph, it must de propagated correctly to all succeeding nodes;
* - It handles cyclic dependencies correctly (currently only induced by the Memorize_Op).
+ * @param dims: vector of vector of dimensions of the inputs of the graphView
* @return bool Whether it succeeded or not (failure can either raise an exception or return false)
*/
bool forwardDims(const std::vector<std::vector<DimSize_t>>& dims = {}, bool allowDataDependency = false);
diff --git a/python_binding/graph/pybind_GraphView.cpp b/python_binding/graph/pybind_GraphView.cpp
index 725cde2b3..291d4400e 100644
--- a/python_binding/graph/pybind_GraphView.cpp
+++ b/python_binding/graph/pybind_GraphView.cpp
@@ -241,7 +241,7 @@ void init_GraphView(py::module& m) {
.def("get_nodes", &GraphView::getNodes,
R"mydelimiter(
Get the Nodes in the GraphView.
-
+
:return: List of the GraphView's Nodes
:rtype: List[Node]
)mydelimiter")
@@ -255,11 +255,70 @@ void init_GraphView(py::module& m) {
:return: The Node of the GraphView with corresponding name
:rtype: Node
)mydelimiter")
- .def("forward_dims", &GraphView::forwardDims, py::arg("dims")=std::vector<std::vector<DimSize_t>>(), py::arg("allow_data_dependency") = false)
- .def("compile", &GraphView::compile, py::arg("backend"), py::arg("datatype"), py::arg("device") = 0, py::arg("dims")=std::vector<std::vector<DimSize_t>>())
- .def("__call__", &GraphView::operator(), py::arg("connectors"))
- .def("set_datatype", &GraphView::setDataType, py::arg("datatype"))
- .def("set_backend", &GraphView::setBackend, py::arg("backend"), py::arg("device") = 0)
+
+ .def("forward_dims", &GraphView::forwardDims, py::arg("dims")=std::vector<std::vector<DimSize_t>>(), py::arg("allow_data_dependency") = false,
+ R"mydelimiter(
+ Compute dimensions of input/output Tensors for each Node's
+ Operator of the GraphView.
+ This function verifies the following conditions:
+ - Every node's dimensions will be updated regardless of if dims were previously forwarded or not;
+ - Updates are made in node dependencies order, because if dims have changed
+ at any point in the graph, it must de propagated correctly to all succeeding nodes;
+ - It handles cyclic dependencies correctly (currently only induced by the Memorize_Op).
+ :param dims: List of list of dimensions of the inputs of the graphView
+ :type dims: List[List[int]]
+ :return: Whether it succeeded or not (failure can either raise an exception or return false)
+ :rtype: bool
+ )mydelimiter")
+
+ .def("compile", &GraphView::compile, py::arg("backend"), py::arg("datatype"), py::arg("device") = 0, py::arg("dims")=std::vector<std::vector<DimSize_t>>(),
+ R"mydelimiter(
+ Assert Datatype, Backend, data format and dimensions along the GraphView are coherent.
+ If not, apply the required transformations.
+ Sets the GraphView ready for computation in four steps:
+ 1 - Assert input Tensors' datatype is compatible with each Operator's datatype.
+ If not, a conversion Operator is inserted.
+ 2 - Assert input Tensors' backend is compatible with each Operator's backend.
+ If not, add a Transmitter Operator.
+ 3 - Assert data format (NCHW, NHWC, ...) of each Operator's input Tensor is
+ compatible with the selected kernel.
+ If not, add a Transpose Operator.
+ 4 - Propagate Tensor dimensions through the consecutive Operators.
+ :param backend: Backend used, default is cpu
+ :type backend: string
+ :param datatype: Aidge Datatype used, default is float32
+ :type datatype: DataType
+ :param device: Device to be set
+ :type device: int
+ :param dims: List of list of dimensions of the inputs of the graphView
+ :type dims: List[List[int]]
+ )mydelimiter")
+
+ .def("__call__", &GraphView::operator(), py::arg("connectors"),
+ R"mydelimiter(
+ Functional operator for user-friendly connection interface using an ordered set of Connectors.
+ :parm connectors: The connector be added to current connection
+ :type connectors: Connector
+ :return: The new connector
+ :rtype: Connector
+ )mydelimiter")
+
+ .def("set_datatype", &GraphView::setDataType, py::arg("datatype"),
+ R"mydelimiter(
+ Set the same data type for each Operator of the GraphView object's Nodes.
+ :param datatype: DataType to be set
+ :type datatype: DataType
+ )mydelimiter")
+
+ .def("set_backend", &GraphView::setBackend, py::arg("backend"), py::arg("device") = 0,
+ R"mydelimiter(
+ Set the same backend for each Operator of the GraphView object's Nodes.
+ :param backend: Backen used, default is cpu
+ :type backend: string
+ :param device: Device to be set
+ :type device: int
+ )mydelimiter")
+
.def("get_ordered_nodes", &GraphView::getOrderedNodes, py::arg("reversed") = false,
R"mydelimiter(
Get a topological node order for an acyclic walk of the graph.
--
GitLab