Add sphinx doc in PyBind file. Provide more documentation

docs/api/visualdl.rst

@@ -2,5 +2,13 @@
 API
 ==========
 
+Python API
+----------
+
 .. automodule:: visualdl
-    :members: LogReader, LogWriter
+    :members: LogReader, LogWriter
+
+PyBind Wrapper Classes
+-----------------------
+
+.. automodule:: visualdl.core

visualdl/logic/pybind.cc

@@ -29,7 +29,20 @@ namespace cp = visualdl::components;
   CODE(double);
 
 PYBIND11_MODULE(core, m) {
-  m.doc() = "C++ core of VisualDL";
+  m.doc() = R"pbdoc(
+
+        VisualDL uses PyBind to operate with the c++ framework. Users should use LogWriter to instantiate scalar/histogram/image writer
+
+        .. autoclass:: ScalarWriter__float
+            :members:
+
+        .. autoclass:: HistogramWriter__float
+            :members:
+
+        .. autoclass:: ImageWriter
+            :members:
+
+    )pbdoc";
 
   py::class_<vs::LogReader>(m, "LogReader")
       .def(py::init(
@@ -115,22 +128,63 @@ PYBIND11_MODULE(core, m) {
 #undef ADD_SCALAR_READER
 
 #define ADD_SCALAR_WRITER(T)                          \
-  py::class_<cp::Scalar<T>>(m, "ScalarWriter__" #T)   \
+  py::class_<cp::Scalar<T>>(m, "ScalarWriter__" #T, R"pbdoc(
+        PyBind class. Must instantiate through the LogWriter.
+      )pbdoc")                                        \
       .def("set_caption", &cp::Scalar<T>::SetCaption) \
-      .def("add_record", &cp::Scalar<T>::AddRecord);
+      .def("add_record", &cp::Scalar<T>::AddRecord, R"pbdoc(
+        add a record with the step and value
+
+        :param step: This value appears at this step in the run.
+        :type step: integer
+        :param value: The scalar value to be recorded.
+        :type value: float
+      )pbdoc");
   ADD_SCALAR_WRITER(int);
   ADD_SCALAR_WRITER(float);
   ADD_SCALAR_WRITER(double);
 #undef ADD_SCALAR_WRITER
 
   // clang-format on
-  py::class_<cp::Image>(m, "ImageWriter")
-      .def("set_caption", &cp::Image::SetCaption)
-      .def("start_sampling", &cp::Image::StartSampling)
-      .def("is_sample_taken", &cp::Image::IsSampleTaken)
-      .def("finish_sampling", &cp::Image::FinishSampling)
-      .def("set_sample", &cp::Image::SetSample)
-      .def("add_sample", &cp::Image::AddSample);
+  py::class_<cp::Image>(m, "ImageWriter", R"pbdoc(
+        PyBind class. Must instantiate through the LogWriter.
+      )pbdoc")
+      .def("set_caption", &cp::Image::SetCaption, R"pbdoc(
+        PyBind class. Must instantiate through the LogWriter.
+      )pbdoc")
+      .def("start_sampling", &cp::Image::StartSampling, R"pbdoc(
+        Start a sampling period, this interface will start a new reservoir sampling phase.
+      )pbdoc")
+      .def("is_sample_taken", &cp::Image::IsSampleTaken, R"pbdoc(
+        Will this sample be taken, this interface is introduced to reduce the cost
+        of copy image data, by testing whether this image will be sampled, and only
+        copy data when it should be sampled. In that way, most of un-sampled image
+        data need not be copied or processed at all.
+
+        :return: Index
+        :rtype: integer
+              )pbdoc")
+      .def("finish_sampling", &cp::Image::FinishSampling, R"pbdoc(
+        End a sampling period, it will clear all states for reservoir sampling.
+      )pbdoc")
+      .def("set_sample", &cp::Image::SetSample, R"pbdoc(
+        Store the image shape with the flatten image data.
+
+        :param index:
+        :type index: integer
+        :param image_shape: Image size
+        :type image_shape: tuple
+        :param image_data: Flatten image data
+        :type image_data: list
+              )pbdoc")
+      .def("add_sample", &cp::Image::AddSample, R"pbdoc(
+        A combined interface for is_sample_taken and set_sample, simpler but is less efficient.
+
+        :param image_shape: Image size
+        :type image_shape: tuple
+        :param image_data: Flatten image data
+        :type image_data: list
+              )pbdoc");
 
   py::class_<cp::ImageReader::ImageRecord>(m, "ImageRecord")
       // TODO(ChunweiYan) make these copyless.
@@ -147,9 +201,19 @@ PYBIND11_MODULE(core, m) {
       .def("record", &cp::ImageReader::record)
       .def("timestamp", &cp::ImageReader::timestamp);
 
-#define ADD_HISTOGRAM_WRITER(T)                           \
-  py::class_<cp::Histogram<T>>(m, "HistogramWriter__" #T) \
-      .def("add_record", &cp::Histogram<T>::AddRecord);
+#define ADD_HISTOGRAM_WRITER(T) \
+  py::class_<cp::Histogram<T>>(m, "HistogramWriter__" #T, R"pbdoc(
+        PyBind class. Must instantiate through the LogWriter.
+      )pbdoc")                  \
+      .def("add_record", &cp::Histogram<T>::AddRecord, R"pbdoc(
+        add a record with the step and histogram_value
+
+        :param step: This value appears at this step in the run.
+        :type step: integer
+        :param histogram_value: A flatten list of the distribution value. EX: [1, 2, 100, 2, 3, 200] will draw a histogram where
+         the value between 1~2 is 100 and the value between 2~3 is 200
+        :type histogram_value: list
+              )pbdoc");
   ADD_FULL_TYPE_IMPL(ADD_HISTOGRAM_WRITER)
 #undef ADD_HISTOGRAM_WRITER
 

visualdl/python/storage.py

@@ -168,6 +168,8 @@ def scalar(self, tag, type='float'):
 
         :param tag: The scalar writer will label the data with tag
         :type tag: basestring
+        :return: A scalar writer to handle step and value records
+        :rtype: ScalarWriter
         """
         check_tag_name_valid(tag)
         type2scalar = {
@@ -183,6 +185,12 @@ def image(self, tag, num_samples, step_cycle=1):
 
         :param tag: The image writer will label the image with tag
         :type tag: basestring
+        :param num_samples: how many samples to take in a step.
+        :type num_samples: integer
+        :param step_cycle: store every `step_cycle` as a record.
+        :type step_cycle: integer
+        :return: A image writer to sample images
+        :rtype: ImageWriter
         """
         check_tag_name_valid(tag)
         return self.writer.new_image(tag, num_samples, step_cycle)
@@ -194,6 +202,8 @@ def histogram(self, tag, num_buckets, type='float'):
 
         :param tag: The histogram writer will label the data with tag
         :type tag: basestring
+        :return: A histogram writer to record distribution
+        :rtype: HistogramWriter
         """
         check_tag_name_valid(tag)
         types = {