Init sphinx documentation for C++ API

Summary:
This commit uses sphinx to generate C++ API documentation
whose style and theme are consistent with the Python API.

docker/ubuntu-16.04-cpu/Dockerfile

 FROM ubuntu:16.04
 
 RUN \
-  apt-get update && apt-get install -y --no-install-recommends \
+  apt-get update && apt-get install -y \
+  --no-install-recommends \
+  --allow-change-held-packages \
   build-essential \
   cmake \
   git \

docker/ubuntu-16.04-cuda10.0-cudnn7/Dockerfile

@@ -2,7 +2,9 @@ FROM nvidia/cuda:10.0-cudnn7-devel-ubuntu16.04
 
 RUN \
   rm /etc/apt/sources.list.d/cuda.list && \
-  apt-get update && apt-get install -y --no-install-recommends \
+  apt-get update && apt-get install -y \
+  --no-install-recommends \
+  --allow-change-held-packages \
   build-essential \
   cmake \
   git \

docs/README.md

@@ -3,27 +3,43 @@ Building Dragon Documentation
 
 This page will help you to build the following documentations:
 
-Dragon C++ API: https://dragon.seetatech.com/api/cc
+Python API: https://dragon.seetatech.com/api/python
 
-Dragon Python API: https://dragon.seetatech.com/api/python
+C++ API: https://dragon.seetatech.com/api/cc
 
-Build Documentation of C++ API
-------------------------------
+Requirements
+------------
+
+- sphinx >= 3.0.2
 
 ```bash
-cd dragon/docs/api/cc
-doxygen Doxyfile
+pip install sphinx
 ```
 
-Then, open the ```docs/api/cc/html/index.html``` in your browser.
+- sphinx_seeta_theme
+
+```bash
+pip install sphinx_seeta_theme
+```
+
+- doxygen (C++ API only)
+
+See: http://www.doxygen.org/download.html
 
 Build Documentation of Python API
 ---------------------------------
 
 ```bash
-pip install sphinx_seeta_theme
-cd dragon/docs/api/python
-make html
+cd dragon/docs/api/python && make html
+```
+
+Then, open the ``docs/_build/api/python/index.html`` in your browser.
+
+Build Documentation of C++ API
+------------------------------
+
+```bash
+cd dragon/docs/api/cc && make doxygen && make html
 ```
 
-Then, open the ```docs/api/python/index.html``` in your browser.
+Then, open the ``docs/_build/api/cc/index.html`` in your browser.

docs/api/cc/Doxyfile

@@ -32,7 +32,7 @@ DOXYFILE_ENCODING      = UTF-8
 # title of most generated pages and in a few other places.
 # The default value is: My Project.
 
-PROJECT_NAME           = "Dragon - C++ API"
+PROJECT_NAME           =
 
 # The PROJECT_NUMBER tag can be used to enter a project or revision number. This
 # could be handy for archiving the generated documentation or if some version
@@ -44,21 +44,21 @@ PROJECT_NUMBER         =
 # for a project that appears at the top of each page and should give viewer a
 # quick idea about the purpose of the project. Keep the description short.
 
-PROJECT_BRIEF          = "A Computation Graph Virtual Machine Based Deep Learning Framework"
+PROJECT_BRIEF          =
 
 # With the PROJECT_LOGO tag one can specify a logo or an icon that is included
 # in the documentation. The maximum height of the logo should not exceed 55
 # pixels and the maximum width should not exceed 200 pixels. Doxygen will copy
 # the logo to the output directory.
 
-PROJECT_LOGO           = images/logo.png
+PROJECT_LOGO           =
 
 # The OUTPUT_DIRECTORY tag is used to specify the (relative or absolute) path
 # into which the generated documentation will be written. If a relative path is
 # entered, it will be relative to the location where doxygen was started. If
 # left blank the current directory will be used.
 
-OUTPUT_DIRECTORY       = ""
+OUTPUT_DIRECTORY       = "../../_build/api/cc_doxygen"
 
 # If the CREATE_SUBDIRS tag is set to YES then doxygen will create 4096 sub-
 # directories (in 2 levels) under the output directory of each output format and
@@ -143,7 +143,7 @@ ALWAYS_DETAILED_SEC    = NO
 # operators of the base classes will not be shown.
 # The default value is: NO.
 
-INLINE_INHERITED_MEMB  = NO
+INLINE_INHERITED_MEMB  = YES
 
 # If the FULL_PATH_NAMES tag is set to YES, doxygen will prepend the full path
 # before files name in the file list and in the header files. If set to NO the
@@ -1044,7 +1044,7 @@ VERBATIM_HEADERS       = YES
 # generated with the -Duse-libclang=ON option for CMake.
 # The default value is: NO.
 
-CLANG_ASSISTED_PARSING = NO
+# CLANG_ASSISTED_PARSING = NO
 
 # If clang assisted parsing is enabled you can provide the compiler with command
 # line options that you would normally use when invoking the compiler. Note that
@@ -1052,7 +1052,7 @@ CLANG_ASSISTED_PARSING = NO
 # specified with INPUT and INCLUDE_PATH.
 # This tag requires that the tag CLANG_ASSISTED_PARSING is set to YES.
 
-CLANG_OPTIONS          =
+# CLANG_OPTIONS          =
 
 # If clang assisted parsing is enabled you can provide the clang parser with the
 # path to the compilation database (see:
@@ -1063,7 +1063,7 @@ CLANG_OPTIONS          =
 # generated with the -Duse-libclang=ON option for CMake.
 # The default value is: 0.
 
-CLANG_COMPILATION_DATABASE_PATH                                        = 0
+# CLANG_COMPILATION_DATABASE_PATH                                        = 0
 
 #---------------------------------------------------------------------------
 # Configuration options related to the alphabetical class index
@@ -1098,7 +1098,7 @@ IGNORE_PREFIX          =
 # If the GENERATE_HTML tag is set to YES, doxygen will generate HTML output
 # The default value is: YES.
 
-GENERATE_HTML          = YES
+GENERATE_HTML          = NO
 
 # The HTML_OUTPUT tag is used to specify where the HTML docs will be put. If a
 # relative path is entered the value of OUTPUT_DIRECTORY will be put in front of
@@ -1930,7 +1930,7 @@ MAN_LINKS              = NO
 # captures the structure of the code including all documentation.
 # The default value is: NO.
 
-GENERATE_XML           = NO
+GENERATE_XML           = YES
 
 # The XML_OUTPUT tag is used to specify where the XML pages will be put. If a
 # relative path is entered the value of OUTPUT_DIRECTORY will be put in front of
@@ -2083,9 +2083,7 @@ INCLUDE_FILE_PATTERNS  =
 # recursively expanded use the := operator instead of the = operator.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-PREDEFINED             = WITH_MPI \
-                         WITH_CUDA \
-                         WITH_CUDNN \
+PREDEFINED             = DRAGON_API= USE_MPI USE_CUDA USE_CUDNN USE_NCCL
 
 # If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this
 # tag can be used to specify a list of macro names that should be expanded. The

docs/api/cc/Makefile

+# Makefile for Sphinx documentation
+# You can set these variables from the command line.
+SPHINXOPTS    =
+SPHINXBUILD   = sphinx-build
+PAPER         =
+BUILDDIR      = ../../_build/api/cc
+
+# User-friendly check for sphinx-build
+ifeq ($(shell which $(SPHINXBUILD) >/dev/null 2>&1; echo $$?), 1)
+$(error The '$(SPHINXBUILD)' command was not found. Make sure you have Sphinx installed.)
+endif
+
+# Internal variables.
+PAPEROPT_a4     = -D latex_paper_size=a4
+PAPEROPT_letter = -D latex_paper_size=letter
+ALLSPHINXOPTS   = -d $(BUILDDIR)/doctrees $(PAPEROPT_$(PAPER)) $(SPHINXOPTS) .
+NUMBER_OF_PROCESSORS:=$(shell getconf _NPROCESSORS_ONLN)
+
+.PHONY: help clean html latex latexpdf
+
+help:
+	@echo "Please use \`make <target>' where <target> is one of"
+	@echo "  doxygen    to make Doxygen XML files"
+	@echo "  html       to make standalone HTML files"
+	@echo "  latex      to make LaTeX files, you can set PAPER=a4 or PAPER=letter"
+	@echo "  latexpdf   to make LaTeX files and run them through pdflatex"
+
+clean:
+	rm -rf $(BUILDDIR)/*
+
+doxygen:
+	mkdir -p $(BUILDDIR)_doxygen && doxygen
+	@echo
+	@echo "Build finished. The Doxygen XML files are in $(BUILDDIR)_doxygen/xml."
+
+html:
+	$(SPHINXBUILD) -b html -j ${NUMBER_OF_PROCESSORS} $(ALLSPHINXOPTS) $(BUILDDIR)
+	@echo
+	@echo "Build finished. The HTML pages are in $(BUILDDIR)."
+
+latex:
+	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)-latex
+	@echo
+	@echo "Build finished; the LaTeX files are in $(BUILDDIR)-latex."
+	@echo "Run \`make' in that directory to run these through (pdf)latex" \
+	      "(use \`make latexpdf' here to do that automatically)."
+
+latexpdf:
+	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)-latex
+	@echo "Running LaTeX files through pdflatex..."
+	$(MAKE) -C $(BUILDDIR)/latex all-pdf
+	@echo "pdflatex finished; the PDF files are in $(BUILDDIR)-latex."

docs/api/cc/conf.py

+# ------------------------------------------------------------
+# Copyright (c) 2017-present, SeetaTech, Co.,Ltd.
+#
+# Licensed under the BSD 2-Clause License.
+# You should have received a copy of the BSD 2-Clause License
+# along with the software. If not, See,
+#
+#    <https://opensource.org/licenses/BSD-2-Clause>
+#
+# ------------------------------------------------------------
+"""Sphinx configuration for C++ API."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from sphinx_seeta_theme import HTMLTranslator
+from sphinx_seeta_theme import HTMLTranslatorV2
+from sphinx_seeta_theme import setup as setup_v1
+
+
+def path_to(href, index=False):
+    if index:
+        if len(href) == 0:
+            return 'index.html'
+        return href + '/index.html'
+    else:
+        return href + '.html'
+
+
+# Basic
+html_static_path = ['../_static']
+exclude_patterns = ['../_build']
+master_doc = 'index'
+source_suffix = '.rst'
+
+# Extension
+extensions = [
+    'sphinx.ext.autodoc',
+    'sphinx.ext.viewcode',
+    'sphinx.ext.napoleon',
+    'sphinxcontrib.katex',
+    'breathe',
+]
+napoleon_use_rtype = False
+
+# Project
+project = 'dragon'
+copyright = 'Copyright (c) 2017-present, SeetaTech, Co.,Ltd'
+author = 'SeetaTech'
+with open('../../../dragon/version.txt', 'r') as f:
+    version = f.read().strip()
+
+# Sphinx
+c_id_attributes = ['DRAGON_API']
+cpp_id_attributes = ['DRAGON_API']
+
+# Breathe
+breathe_projects = {'dragon': '../../_build/api/cc_doxygen/xml/'}
+breathe_default_project = 'dragon'
+
+# HTML
+html_theme = 'seeta'
+html_title = ''
+html_short_title = ''
+html_logo = '../_static/images/dragon.png'
+html_favicon = '../_static/favicon.ico'
+html_copy_source = False
+html_show_sourcelink = False
+html_show_sphinx = False
+html_show_copyright = False
+html_scaled_image_link = False
+html_theme_options = {
+    'navbar_links': {
+        'Install': path_to('../../install', 1),
+        'API': [
+            ('master', path_to('../../api/python', 1)),
+            ('versions...', path_to('../../versions', 1)),
+        ],
+        'Github': 'https://github.com/seetaresearch/dragon',
+    },
+    'navbar_logo_link': path_to('../..', 1),
+    'sidebar_title': 'C++ v{}'.format(version),
+    'sidebar_title_link': path_to('../../versions', 1),
+    'breadcrumb_links': [
+        ('Dragon', path_to('../..', 1)),
+        ('API', path_to('../../versions', 1)),
+        ('Dragon v{}'.format(version.replace('a0', '-a0')), path_to('../../api', 1)),
+        ('C++', path_to('', 1)),
+    ],
+}
+html_sidebars = {
+    'index': ['localtoc.html'],
+    'dragon': ['localtoc.html'],
+    'dragon/**': ['localtoc.html'],
+    '_modules/**': ['localtoc.html'],
+    'search': ['localtoc.html'],
+}
+
+# LaTex
+latex_documents = [(
+    master_doc,
+    'dragon.tex',
+    'Dragon - C++ API',
+    author,
+    'manual',
+)]
+latex_elements = {
+    'utf8extra': '',
+    'inputenc': '',
+    'babel': r'''\usepackage[english]{babel}''',
+    'preamble': r'''
+\usepackage{enumitem}
+\usepackage{tocloft}
+\renewcommand{\cfttoctitlefont}{\huge\bfseries}
+\usepackage{fontspec}
+\setmainfont{Source Serif Pro}
+\setsansfont{Source Serif Pro}
+\setmonofont{Source Serif Pro}
+\setcounter{tocdepth}{2}
+\usepackage[draft]{minted}
+\fvset{breaklines=true, breakanywhere=true}
+\setlength{\headheight}{13.6pt}
+\setlength{\itemindent}{-1pt}
+\makeatletter
+    \renewcommand*\l@subsection{\@dottedtocline{2}{3.8em}{3.8em}}
+    \fancypagestyle{normal}{
+        \fancyhf{}
+        \fancyfoot[LE,RO]{{\py@HeaderFamily\thepage}}
+        \fancyfoot[LO]{{\py@HeaderFamily\nouppercase{\rightmark}}}
+        \fancyfoot[RE]{{\py@HeaderFamily\nouppercase{\leftmark}}}
+        \fancyhead[LE,RO]{{\py@HeaderFamily}}
+     }
+\makeatother
+''',
+    'maketitle': r'''
+\pagenumbering{Roman} %% % to avoid page 1 conflict with actual page 1
+
+\makeatletter
+\begin{titlepage}
+
+    \noindent\rule[0.25\baselineskip]{\textwidth}{1pt}
+
+    \vspace*{5mm}
+    \begin{figure}[!h]
+        \raggedleft
+        \includegraphics[scale=0.3]{logo.png}
+    \end{figure}
+
+    \raggedleft
+    \vspace*{5mm}
+    \textbf{\Huge \@title}
+
+    \vspace*{40mm}
+    \LARGE \@author
+
+\end{titlepage}
+\makeatother
+
+\pagenumbering{arabic}
+''',
+    'pointsize': '10pt',
+    'figure_align': 'H',
+    'printindex': '',
+    'sphinxsetup': ' \
+        hmargin={0.75in,0.75in}, \
+        vmargin={0.5in,1in}, \
+        verbatimhintsturnover=false, \
+        verbatimsep=0.75em, \
+        verbatimhintsturnover=false, \
+        verbatimwithframe=false, \
+        VerbatimColor={rgb}{0.949,0.949,0.949}, \
+        HeaderFamily=\\rmfamily\\bfseries',
+}
+
+latex_domain_indices = False
+latex_engine = 'xelatex'
+latex_logo = '../_static/images/logo.png'
+
+
+# Application API
+class HTMLTranslatorV3(HTMLTranslatorV2):
+    """Custom html translator."""
+
+    def depart_desc_content(self, node):
+        """Remove the sub classees."""
+        HTMLTranslatorV2.depart_desc_content(self, node)
+        para_start, para_end = -1, -1
+        for i, text in enumerate(self.body):
+            if para_start > 0 and text.startswith('</p>'):
+                para_end = i
+                break
+            if text.startswith('<p>') and \
+                    self.body[i + 1].startswith('Subclassed by'):
+                para_start = i
+        if para_start > 0 and para_end > 0:
+            self.body = self.body[:para_start] + self.body[para_end + 1:]
+
+    def depart_desc_parameterlist(self, node):
+        """Remove the trailing newline to match the google c++ style."""
+        HTMLTranslator.depart_desc_parameterlist(self, node)
+
+
+def setup(app):
+    """Custom application setup."""
+    return setup_v1(app, HTMLTranslatorV3)

docs/api/cc/dragon/core.rst

+dragon/core
+===========
+
+.. only:: html
+
+  Classes
+  -------
+
+  `class CPUContext <core/CPUContext.html>`_
+  : The cpu device context.
+
+  `class CUDAContext <core/CPUContext.html>`_
+  : The cuda device context.
+
+  `class Graph <core/Graph.html>`_
+  : Graph to execute operators sequentially.
+
+  `class Operator <core/Operator.html>`_
+  : The base operator class with context.
+
+  `class Tensor <core/Tensor.html>`_
+  : The base tensor class, manage memory or not.
+
+  `class TypeMeta <core/TypeMeta.html>`_
+  : Metaclass for all types.
+
+  `class UnifiedMemory <core/UnifiedMemory.html>`_
+  : Memory to manage both the host and device data.
+
+  `class Workspace <core/Workspace.html>`_
+  : Sandbox to isolate the resources and computations.
+
+.. toctree::
+  :hidden:
+
+  core/CPUContext
+  core/CUDAContext
+  core/Graph
+  core/Operator
+  core/Tensor
+  core/TypeMeta
+  core/UnifiedMemory
+  core/Workspace
+
+.. raw:: html
+
+  <style>
+  h1:before {
+    content: "Routine: ";
+    color: #103d3e;
+  }
+  </style>

docs/api/cc/dragon/core/CPUContext.rst

+CPUContext
+==========
+
+.. doxygenclass:: dragon::CPUContext
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::CPUContext::CPUContext()
+.. doxygenfunction:: dragon::CPUContext::CPUContext(unsigned int random_seed)
+.. doxygenfunction:: dragon::CPUContext::CPUContext(const DeviceOption &option)
+
+Public Functions
+----------------
+
+Copy
+####
+.. doxygenfunction:: dragon::CPUContext::Copy
+
+Delete
+######
+.. doxygenfunction:: dragon::CPUContext::Delete
+
+FinishDeviceComputation
+#######################
+.. doxygenfunction:: dragon::CPUContext::FinishDeviceComputation
+
+Memset
+######
+.. doxygenfunction:: dragon::CPUContext::Memset
+
+MemsetAsync
+###########
+.. doxygenfunction:: dragon::CPUContext::MemsetAsync
+
+Memcpy
+######
+.. doxygenfunction:: dragon::CPUContext::Memcpy
+
+MemcpyAsync
+###########
+.. doxygenfunction:: dragon::CPUContext::MemcpyAsync
+
+New
+###
+.. doxygenfunction:: dragon::CPUContext::New
+
+SwitchToDevice
+##############
+.. doxygenfunction:: dragon::CPUContext::SwitchToDevice()
+
+SwitchToDevice
+##############
+.. doxygenfunction:: dragon::CPUContext::SwitchToDevice(int stream)
+
+device
+######
+.. doxygenfunction:: dragon::CPUContext::device
+
+rand_generator
+##############
+.. doxygenfunction:: dragon::CPUContext::rand_generator
+
+set_stream
+##########
+.. doxygenfunction:: dragon::CPUContext::set_stream
+
+stream
+######
+.. doxygenfunction:: dragon::CPUContext::stream
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/CUDAContext.rst

+CUDAContext
+===========
+
+.. doxygenclass:: dragon::CUDAContext
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::CUDAContext::CUDAContext()
+.. doxygenfunction:: dragon::CUDAContext::CUDAContext(int device)
+.. doxygenfunction:: dragon::CUDAContext::CUDAContext(const DeviceOption &option)
+
+Public Functions
+----------------
+
+Copy
+####
+.. doxygenfunction:: dragon::CUDAContext::Copy
+
+Delete
+######
+.. doxygenfunction:: dragon::CUDAContext::Delete
+
+FinishDeviceComputation
+#######################
+.. doxygenfunction:: dragon::CUDAContext::FinishDeviceComputation
+
+Memset
+######
+.. doxygenfunction:: dragon::CUDAContext::Memset
+
+MemsetAsync
+###########
+.. doxygenfunction:: dragon::CUDAContext::MemsetAsync
+
+Memcpy
+######
+.. doxygenfunction:: dragon::CUDAContext::Memcpy(size_t n, void *dest, const void *src)
+
+Memcpy
+######
+.. doxygenfunction:: dragon::CUDAContext::Memcpy(size_t n, void *dest, const void *src, int device)
+
+MemcpyAsync
+###########
+.. doxygenfunction:: dragon::CUDAContext::MemcpyAsync
+
+New
+###
+.. doxygenfunction:: dragon::CUDAContext::New
+
+SwitchToDevice
+##############
+.. doxygenfunction:: dragon::CUDAContext::SwitchToDevice()
+
+SwitchToDevice
+##############
+.. doxygenfunction:: dragon::CUDAContext::SwitchToDevice(int stream)
+
+SynchronizeStream
+#################
+.. doxygenfunction:: dragon::CUDAContext::SynchronizeStream
+
+cublas_handle
+#############
+.. doxygenfunction:: dragon::CUDAContext::cublas_handle
+
+cuda_stream
+###########
+.. doxygenfunction:: dragon::CUDAContext::cuda_stream()
+
+cuda_stream
+###########
+.. doxygenfunction:: dragon::CUDAContext::cuda_stream(int device, int stream)
+
+cudnn_handle
+############
+.. doxygenfunction:: dragon::CUDAContext::cudnn_handle
+
+curand_generator
+################
+.. doxygenfunction:: dragon::CUDAContext::curand_generator
+
+rand_generator
+##############
+.. doxygenfunction:: dragon::CUDAContext::rand_generator
+
+device
+######
+.. doxygenfunction:: dragon::CUDAContext::device
+
+set_stream
+##########
+.. doxygenfunction:: dragon::CUDAContext::set_stream
+
+stream
+######
+.. doxygenfunction:: dragon::CUDAContext::stream
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/Graph.rst

+Graph
+=====
+
+.. doxygenclass:: dragon::Graph
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::Graph::Graph(const GraphDef& def, Workspace* ws)
+
+Public Functions
+----------------
+
+Create
+######
+.. doxygenfunction:: dragon::Graph::Create
+
+Run
+###
+.. doxygenfunction:: dragon::Graph::Run
+
+arg
+###
+.. doxygenfunction:: dragon::Graph::arg
+
+args
+####
+.. doxygenfunction:: dragon::Graph::args
+
+def
+###
+.. doxygenfunction:: dragon::Graph::def
+
+optimized_def
+#############
+.. doxygenfunction:: dragon::Graph::optimized_def
+
+name
+####
+.. doxygenfunction:: dragon::Graph::name
+
+phase
+#####
+.. doxygenfunction:: dragon::Graph::phase
+
+ws
+##
+.. doxygenfunction:: dragon::Graph::ws
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/Operator.rst

+Operator
+========
+
+.. doxygenclass:: dragon::Operator
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::Operator::Operator(const OperatorDef &def, Workspace *ws)
+
+Public Functions
+----------------
+
+Arg
+###
+.. doxygenfunction:: dragon::Operator::Arg
+
+Args
+####
+.. doxygenfunction:: dragon::Operator::Args
+
+Buffer
+######
+.. doxygenfunction:: dragon::Operator::Buffer
+
+Fuse
+####
+.. doxygenfunction:: dragon::Operator::Fuse
+
+Input
+#####
+.. doxygenfunction:: dragon::Operator::Input
+
+InputSize
+#########
+.. doxygenfunction:: dragon::Operator::InputSize
+
+Output
+######
+.. doxygenfunction:: dragon::Operator::Output(int i)
+
+MessageForUnsupported
+#####################
+.. doxygenfunction:: dragon::Operator::MessageForUnsupported
+
+Output
+######
+.. doxygenfunction:: dragon::Operator::Output(int i, const vec32_t &inputs)
+
+OutputSize
+##########
+.. doxygenfunction:: dragon::Operator::OutputSize
+
+Run
+###
+.. doxygenfunction:: dragon::Operator::Run
+
+UpdateFrom
+##########
+.. doxygenfunction:: dragon::Operator::UpdateFrom
+
+data_format
+###########
+.. doxygenfunction:: dragon::Operator::data_format
+
+arg
+###
+.. doxygenfunction:: dragon::Operator::arg
+
+args
+####
+.. doxygenfunction:: dragon::Operator::args
+
+def
+###
+.. doxygenfunction:: dragon::Operator::def
+
+dtype
+#####
+.. doxygenfunction:: dragon::Operator::dtype
+
+handle
+######
+.. doxygenfunction:: dragon::Operator::handle
+
+name
+####
+.. doxygenfunction:: dragon::Operator::name
+
+type
+####
+.. doxygenfunction:: dragon::Operator::type
+
+phase
+#####
+.. doxygenfunction:: dragon::Operator::phase
+
+ws
+##
+.. doxygenfunction:: dragon::Operator::ws
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/Tensor.rst

+Tensor
+======
+
+.. doxygenclass:: dragon::Tensor
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::Tensor::Tensor()
+.. doxygenfunction:: dragon::Tensor::Tensor(const string &name)
+.. doxygenfunction:: dragon::Tensor::Tensor(const vec64_t &dims)
+.. doxygenfunction:: dragon::Tensor::Tensor(const vec32_t &dims)
+.. doxygenfunction:: dragon::Tensor::Tensor(const TypeMeta &meta)
+
+Public Functions
+----------------
+
+CopyFrom
+########
+.. doxygenfunction:: dragon::Tensor::CopyFrom(const Tensor &other, Context *ctx)
+
+CopyFrom
+########
+.. doxygenfunction:: dragon::Tensor::CopyFrom(const vector<VectorType> &other)
+
+CopyTo
+######
+.. doxygenfunction:: dragon::Tensor::CopyTo
+
+DimString
+#########
+.. doxygenfunction:: dragon::Tensor::DimString() const
+
+DimString
+#########
+.. doxygenfunction:: dragon::Tensor::DimString(const vector<int64_t> &dims)
+
+IsType
+######
+.. doxygenfunction:: dragon::Tensor::IsType
+
+Reset
+#####
+.. doxygenfunction:: dragon::Tensor::Reset
+
+Reshape
+#######
+.. doxygenfunction:: dragon::Tensor::Reshape
+
+ReshapeLike
+###########
+.. doxygenfunction:: dragon::Tensor::ReshapeLike
+
+Share
+#####
+.. doxygenfunction:: dragon::Tensor::Share
+
+SwitchToDevice
+##############
+.. doxygenfunction:: dragon::Tensor::SwitchToDevice
+
+axis
+####
+.. doxygenfunction:: dragon::Tensor::axis
+
+capacity
+########
+.. doxygenfunction:: dragon::Tensor::capacity
+
+conut
+#####
+.. doxygenfunction:: dragon::Tensor::count() const
+
+conut
+#####
+.. doxygenfunction:: dragon::Tensor::count(int64_t start) const
+
+conut
+#####
+.. doxygenfunction:: dragon::Tensor::count(int64_t start, int64_t end) const
+
+data
+####
+.. doxygenfunction:: dragon::Tensor::data
+
+dim
+###
+.. doxygenfunction:: dragon::Tensor::dim
+
+dims
+####
+.. doxygenfunction:: dragon::Tensor::dims
+
+empty
+#####
+.. doxygenfunction:: dragon::Tensor::empty
+
+has_memory
+##########
+.. doxygenfunction:: dragon::Tensor::has_memory
+
+has_name
+########
+.. doxygenfunction:: dragon::Tensor::has_name
+
+meta
+####
+.. doxygenfunction:: dragon::Tensor::meta
+
+memory
+######
+.. doxygenfunction:: dragon::Tensor::memory
+
+memory_state
+############
+.. doxygenfunction:: dragon::Tensor::memory_state
+
+mutable_data
+############
+.. doxygenfunction:: dragon::Tensor::mutable_data
+
+name
+####
+.. doxygenfunction:: dragon::Tensor::name
+
+nbytes
+######
+.. doxygenfunction:: dragon::Tensor::nbytes
+
+ndim
+####
+.. doxygenfunction:: dragon::Tensor::ndim
+
+raw_data
+########
+.. doxygenfunction:: dragon::Tensor::raw_data
+
+raw_mutable_data
+################
+.. doxygenfunction:: dragon::Tensor::raw_mutable_data()
+
+raw_mutable_data
+################
+.. doxygenfunction:: dragon::Tensor::raw_mutable_data(const TypeMeta &meta)
+
+size
+####
+.. doxygenfunction:: dragon::Tensor::size
+
+stride
+######
+.. doxygenfunction:: dragon::Tensor::stride
+
+strides
+#######
+.. doxygenfunction:: dragon::Tensor::strides
+
+version
+#######
+.. doxygenfunction:: dragon::Tensor::version
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/TypeMeta.rst

+TypeMeta
+========
+
+.. doxygenclass:: dragon::TypeMeta
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::TypeMeta::TypeMeta()
+.. doxygenfunction:: dragon::TypeMeta::TypeMeta(const TypeMeta &src)
+
+Public Functions
+----------------
+
+Copy
+####
+.. doxygenfunction:: dragon::TypeMeta::Copy
+
+Ctor
+####
+.. doxygenfunction:: dragon::TypeMeta::Ctor
+
+Dtor
+####
+.. doxygenfunction:: dragon::TypeMeta::Dtor
+
+Id
+##
+.. doxygenfunction:: dragon::TypeMeta::Id
+
+Itemsize
+########
+.. doxygenfunction:: dragon::TypeMeta::Itemsize
+
+Make
+####
+.. doxygenfunction:: dragon::TypeMeta::Make
+
+Match
+#####
+.. doxygenfunction:: dragon::TypeMeta::Match
+
+copy
+####
+.. doxygenfunction:: dragon::TypeMeta::copy
+
+ctor
+####
+.. doxygenfunction:: dragon::TypeMeta::ctor
+
+dtor
+####
+.. doxygenfunction:: dragon::TypeMeta::dtor
+
+id
+##
+.. doxygenfunction:: dragon::TypeMeta::id
+
+itemsize
+########
+.. doxygenfunction:: dragon::TypeMeta::itemsize
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/UnifiedMemory.rst

+UnifiedMemory
+=============
+
+.. doxygenclass:: dragon::UnifiedMemory
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::UnifiedMemory::UnifiedMemory()
+.. doxygenfunction:: dragon::UnifiedMemory::UnifiedMemory(const TypeMeta &meta, size_t size)
+
+Public Types
+------------
+
+State
+#####
+.. doxygenenum:: dragon::UnifiedMemory::State
+
+Public Functions
+----------------
+
+SwitchToDevice
+##############
+.. doxygenfunction:: dragon::UnifiedMemory::SwitchToDevice
+
+SwitchToCUDADevice
+##################
+.. doxygenfunction:: dragon::UnifiedMemory::SwitchToCUDADevice
+
+ToCPU
+#####
+.. doxygenfunction:: dragon::UnifiedMemory::ToCPU
+
+ToCUDA
+######
+.. doxygenfunction:: dragon::UnifiedMemory::ToCUDA
+
+cpu_data
+########
+.. doxygenfunction:: dragon::UnifiedMemory::cpu_data
+
+cuda_data
+#########
+.. doxygenfunction:: dragon::UnifiedMemory::cuda_data
+
+device
+######
+.. doxygenfunction:: dragon::UnifiedMemory::device
+
+info
+####
+.. doxygenfunction:: dragon::UnifiedMemory::info
+
+mutable_cpu_data
+################
+.. doxygenfunction:: dragon::UnifiedMemory::mutable_cpu_data
+
+mutable_cuda_data
+#################
+.. doxygenfunction:: dragon::UnifiedMemory::mutable_cuda_data
+
+set_cpu_data
+############
+.. doxygenfunction:: dragon::UnifiedMemory::set_cpu_data
+
+set_cuda_data
+#############
+.. doxygenfunction:: dragon::UnifiedMemory::set_cuda_data
+
+size
+####
+.. doxygenfunction:: dragon::UnifiedMemory::size
+
+state
+#####
+.. doxygenfunction:: dragon::UnifiedMemory::state
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/dragon/core/Workspace.rst

+Workspace
+=========
+
+.. doxygenclass:: dragon::Workspace
+
+Constructors
+------------
+
+.. doxygenfunction:: dragon::Workspace::Workspace(const string &name)
+
+Public Functions
+----------------
+
+Clear
+#####
+.. doxygenfunction:: dragon::Workspace::Clear
+
+CreateGraph
+###########
+.. doxygenfunction:: dragon::Workspace::CreateGraph
+
+CreateTensor
+############
+.. doxygenfunction:: dragon::Workspace::CreateTensor
+
+GetFillerInfo
+#############
+.. doxygenfunction:: dragon::Workspace::GetFillerInfo
+
+GetTensor
+#########
+.. doxygenfunction:: dragon::Workspace::GetTensor
+
+HasTensor
+#########
+.. doxygenfunction:: dragon::Workspace::HasTensor
+
+MergeFrom
+#########
+.. doxygenfunction:: dragon::Workspace::MergeFrom
+
+RegisterAlias
+#############
+.. doxygenfunction:: dragon::Workspace::RegisterAlias
+
+ResetTensor
+###########
+.. doxygenfunction:: dragon::Workspace::ResetTensor
+
+RunGraph
+########
+.. doxygenfunction:: dragon::Workspace::RunGraph
+
+RunOperator
+###########
+.. doxygenfunction:: dragon::Workspace::RunOperator
+
+TryGetTensor
+############
+.. doxygenfunction:: dragon::Workspace::TryGetTensor
+
+UniqueName
+##########
+.. doxygenfunction:: dragon::Workspace::UniqueName
+
+data
+####
+.. doxygenfunction:: dragon::Workspace::data(const vector<size_t> &segments)
+
+data
+####
+.. doxygenfunction:: dragon::Workspace::data(const vector<int64_t> &segments)
+
+graphs
+######
+.. doxygenfunction:: dragon::Workspace::graphs
+
+name
+####
+.. doxygenfunction:: dragon::Workspace::name
+
+tensors
+#######
+.. doxygenfunction:: dragon::Workspace::tensors
+
+.. raw:: html
+
+  <style>
+    h1:before {
+      content: "dragon::";
+      color: #103d3e;
+    }
+  </style>

docs/api/cc/index.rst

+Dragon - C++ API
+================
+
+Routines
+--------
+
+.. only:: html
+
+  `Routine core <dragon/core.html>`_
+  : Public API for ``dragon/core`` routine.
+
+.. toctree::
+  :hidden:
+
+  dragon/core

docs/api/cc/make.bat

+:: #########################################################
+:: Command file to build on Windows for Sphinx documentation
+:: #########################################################
+
+@echo off
+
+:: You can set these variables from the command line
+if "%SPHINXBUILD%" == "" (
+	set SPHINXBUILD=sphinx-build
+)
+set BUILDDIR=..\..\_build\api\cc
+set ALLSPHINXOPTS=-d %BUILDDIR%\doctrees %SPHINXOPTS% .
+if NOT "%PAPER%" == "" (
+	set ALLSPHINXOPTS=-D latex_paper_size=%PAPER% %ALLSPHINXOPTS%
+)
+
+if "%1" == "" goto help
+
+if "%1" == "help" (
+	:help
+	echo.Please use `make ^<target^>` where ^<target^> is one of
+  echo.  doxygen    to make Doxygen XML files
+	echo.  html       to make standalone HTML files
+	echo.  debughtml  to make debugging HTML files
+	echo.  latex      to make LaTeX files, you can set PAPER=a4 or PAPER=letter
+	echo.  latexpdf   to make LaTeX files and run them through pdflatex
+	goto end
+)
+
+if "%1" == "clean" (
+	for /d %%i in (%BUILDDIR%\*) do rmdir /q /s %%i
+	del /q /s %BUILDDIR%\*
+	goto end
+)
+
+:: Check if sphinx-build is available and fallback to Python version if any
+%SPHINXBUILD% 2> nul
+if errorlevel 9009 goto sphinx_python
+goto sphinx_ok
+
+:sphinx_python
+
+set SPHINXBUILD=python -m sphinx.__init__
+%SPHINXBUILD% 2> nul
+if errorlevel 9009 (
+	echo.
+	echo.The 'sphinx-build' command was not found. Make sure you have Sphinx
+	echo.installed, then set the SPHINXBUILD environment variable to point
+	echo.to the full path of the 'sphinx-build' executable. Alternatively you
+	echo.may add the Sphinx directory to PATH.
+	echo.
+	echo.If you don't have Sphinx installed, grab it from
+	echo.http://sphinx-doc.org/
+	exit /b 1
+)
+
+:sphinx_ok
+
+if "%1" == "doxygen" (
+	(if exist %BUILDDIR%_doxygen rmdir /q /s %BUILDDIR%_doxygen) && mkdir %BUILDDIR%_doxygen && doxygen
+	if errorlevel 1 exit /b 1
+	echo.
+	echo.Build finished. The Doxygen XML files are in %BUILDDIR%_doxygen/xml.
+	goto end
+)
+
+if "%1" == "html" (
+	%SPHINXBUILD% -b html -j %NUMBER_OF_PROCESSORS% %ALLSPHINXOPTS% %BUILDDIR%
+	if errorlevel 1 exit /b 1
+	echo.
+	echo.Build finished. The HTML pages are in %BUILDDIR%.
+	goto end
+)
+
+if "%1" == "latex" (
+	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%-latex
+	if errorlevel 1 exit /b 1
+	echo.
+	echo.Build finished; the LaTeX files are in %BUILDDIR%-latex.
+	goto end
+)
+
+if "%1" == "latexpdf" (
+	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%-latex
+	cd %BUILDDIR%-latex
+	make all-pdf
+	cd %~dp0
+	echo.
+	echo.Build finished; the PDF files are in %BUILDDIR%-latex.
+	goto end
+)
+
+:end

docs/api/python/Makefile

 # Makefile for Sphinx documentation
-#
 
-# You can set these variables from the command line.
+# You can set these variables from the command line
 SPHINXOPTS    =
 SPHINXBUILD   = sphinx-build
 PAPER         =
-BUILDDIR      = ../../_build/api
+BUILDDIR      = ../../_build/api/python
 
 # User-friendly check for sphinx-build
 ifeq ($(shell which $(SPHINXBUILD) >/dev/null 2>&1; echo $$?), 1)
-$(error The '$(SPHINXBUILD)' command was not found. Make sure you have Sphinx installed, then set the SPHINXBUILD environment variable to point to the full path of the '$(SPHINXBUILD)' executable. Alternatively you can add the directory with the executable to your PATH. If you don't have Sphinx installed, grab it from http://sphinx-doc.org/)
+$(error The '$(SPHINXBUILD)' command was not found. Make sure you have Sphinx installed.)
 endif
 
-# Internal variables.
+# Internal variables
 PAPEROPT_a4     = -D latex_paper_size=a4
 PAPEROPT_letter = -D latex_paper_size=letter
 ALLSPHINXOPTS   = -d $(BUILDDIR)/doctrees $(PAPEROPT_$(PAPER)) $(SPHINXOPTS) .
-# the i18n builder cannot share the environment and doctrees with the others
-I18NSPHINXOPTS  = $(PAPEROPT_$(PAPER)) $(SPHINXOPTS) .
-NPROC:=$(shell getconf _NPROCESSORS_ONLN)
+NUMBER_OF_PROCESSORS:=$(shell getconf _NPROCESSORS_ONLN)
 
-.PHONY: help clean html dirhtml singlehtml pickle json htmlhelp qthelp devhelp epub latex latexpdf text man changes linkcheck doctest coverage gettext
+.PHONY: help clean html debughtml latex latexpdf
 
 help:
 	@echo "Please use \`make <target>' where <target> is one of"
 	@echo "  html       to make standalone HTML files"
-	@echo "  deployhtml to make HTML files copyied to website"
-	@echo "  dirhtml    to make HTML files named index.html in directories"
-	@echo "  singlehtml to make a single large HTML file"
-	@echo "  pickle     to make pickle files"
-	@echo "  json       to make JSON files"
-	@echo "  htmlhelp   to make HTML files and a HTML help project"
-	@echo "  qthelp     to make HTML files and a qthelp project"
-	@echo "  applehelp  to make an Apple Help Book"
-	@echo "  devhelp    to make HTML files and a Devhelp project"
-	@echo "  epub       to make an epub"
+	@echo "  debughtml  to make debugging HTML files"
 	@echo "  latex      to make LaTeX files, you can set PAPER=a4 or PAPER=letter"
 	@echo "  latexpdf   to make LaTeX files and run them through pdflatex"
-	@echo "  latexpdfja to make LaTeX files and run them through platex/dvipdfmx"
-	@echo "  text       to make text files"
-	@echo "  man        to make manual pages"
-	@echo "  texinfo    to make Texinfo files"
-	@echo "  info       to make Texinfo files and run them through makeinfo"
-	@echo "  gettext    to make PO message catalogs"
-	@echo "  changes    to make an overview of all changed/added/deprecated items"
-	@echo "  xml        to make Docutils-native XML files"
-	@echo "  pseudoxml  to make pseudoxml-XML files for display purposes"
-	@echo "  linkcheck  to check all external links for integrity"
-	@echo "  doctest    to run all doctests embedded in the documentation (if enabled)"
-	@echo "  coverage   to run coverage check of the documentation (if enabled)"
 
 clean:
 	rm -rf $(BUILDDIR)/*
 
 html:
-	$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) $(BUILDDIR)/python
+	$(SPHINXBUILD) -b html $(ALLSPHINXOPTS) $(BUILDDIR)
 	@echo
-	@echo "Build finished. The HTML pages are in $(BUILDDIR)/python."
+	@echo "Build finished. The HTML pages are in $(BUILDDIR)."
 
 debughtml:
-	$(SPHINXBUILD) -b html -j ${NPROC} $(ALLSPHINXOPTS) $(BUILDDIR)/python
+	$(SPHINXBUILD) -b html -j ${NUMBER_OF_PROCESSORS} $(ALLSPHINXOPTS) $(BUILDDIR)
 	@echo
-	@echo "Build finished. The HTML pages are in $(BUILDDIR)/python."
-
-dirhtml:
-	$(SPHINXBUILD) -b dirhtml $(ALLSPHINXOPTS) $(BUILDDIR)/dirhtml
-	@echo
-	@echo "Build finished. The HTML pages are in $(BUILDDIR)/dirhtml."
-
-singlehtml:
-	$(SPHINXBUILD) -b singlehtml $(ALLSPHINXOPTS) $(BUILDDIR)/singlehtml
-	@echo
-	@echo "Build finished. The HTML page is in $(BUILDDIR)/singlehtml."
-
-pickle:
-	$(SPHINXBUILD) -b pickle $(ALLSPHINXOPTS) $(BUILDDIR)/pickle
-	@echo
-	@echo "Build finished; now you can process the pickle files."
-
-json:
-	$(SPHINXBUILD) -b json $(ALLSPHINXOPTS) $(BUILDDIR)/json
-	@echo
-	@echo "Build finished; now you can process the JSON files."
-
-htmlhelp:
-	$(SPHINXBUILD) -b htmlhelp $(ALLSPHINXOPTS) $(BUILDDIR)/htmlhelp
-	@echo
-	@echo "Build finished; now you can run HTML Help Workshop with the" \
-	      ".hhp project file in $(BUILDDIR)/htmlhelp."
-
-qthelp:
-	$(SPHINXBUILD) -b qthelp $(ALLSPHINXOPTS) $(BUILDDIR)/qthelp
-	@echo
-	@echo "Build finished; now you can run "qcollectiongenerator" with the" \
-	      ".qhcp project file in $(BUILDDIR)/qthelp, like this:"
-	@echo "# qcollectiongenerator $(BUILDDIR)/qthelp/Dragon.qhcp"
-	@echo "To view the help file:"
-	@echo "# assistant -collectionFile $(BUILDDIR)/qthelp/Dragon.qhc"
-
-applehelp:
-	$(SPHINXBUILD) -b applehelp $(ALLSPHINXOPTS) $(BUILDDIR)/applehelp
-	@echo
-	@echo "Build finished. The help book is in $(BUILDDIR)/applehelp."
-	@echo "N.B. You won't be able to view it unless you put it in" \
-	      "~/Library/Documentation/Help or install it in your application" \
-	      "bundle."
-
-devhelp:
-	$(SPHINXBUILD) -b devhelp $(ALLSPHINXOPTS) $(BUILDDIR)/devhelp
-	@echo
-	@echo "Build finished."
-	@echo "To view the help file:"
-	@echo "# mkdir -p $$HOME/.local/share/devhelp/Dragon"
-	@echo "# ln -s $(BUILDDIR)/devhelp $$HOME/.local/share/devhelp/Dragon"
-	@echo "# devhelp"
-
-epub:
-	$(SPHINXBUILD) -b epub $(ALLSPHINXOPTS) $(BUILDDIR)/epub
-	@echo
-	@echo "Build finished. The epub file is in $(BUILDDIR)/epub."
+	@echo "Build finished. The HTML pages are in $(BUILDDIR)."
 
 latex:
-	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)/latex
+	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)-latex
 	@echo
-	@echo "Build finished; the LaTeX files are in $(BUILDDIR)/latex."
+	@echo "Build finished; the LaTeX files are in $(BUILDDIR)-latex."
 	@echo "Run \`make' in that directory to run these through (pdf)latex" \
 	      "(use \`make latexpdf' here to do that automatically)."
 
 latexpdf:
-	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)/latex
+	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)-latex
 	@echo "Running LaTeX files through pdflatex..."
 	$(MAKE) -C $(BUILDDIR)/latex all-pdf
-	@echo "pdflatex finished; the PDF files are in $(BUILDDIR)/latex."
-
-latexpdfja:
-	$(SPHINXBUILD) -b latex $(ALLSPHINXOPTS) $(BUILDDIR)/latex
-	@echo "Running LaTeX files through platex and dvipdfmx..."
-	$(MAKE) -C $(BUILDDIR)/latex all-pdf-ja
-	@echo "pdflatex finished; the PDF files are in $(BUILDDIR)/latex."
-
-text:
-	$(SPHINXBUILD) -b text $(ALLSPHINXOPTS) $(BUILDDIR)/text
-	@echo
-	@echo "Build finished. The text files are in $(BUILDDIR)/text."
-
-man:
-	$(SPHINXBUILD) -b man $(ALLSPHINXOPTS) $(BUILDDIR)/man
-	@echo
-	@echo "Build finished. The manual pages are in $(BUILDDIR)/man."
-
-texinfo:
-	$(SPHINXBUILD) -b texinfo $(ALLSPHINXOPTS) $(BUILDDIR)/texinfo
-	@echo
-	@echo "Build finished. The Texinfo files are in $(BUILDDIR)/texinfo."
-	@echo "Run \`make' in that directory to run these through makeinfo" \
-	      "(use \`make info' here to do that automatically)."
-
-info:
-	$(SPHINXBUILD) -b texinfo $(ALLSPHINXOPTS) $(BUILDDIR)/texinfo
-	@echo "Running Texinfo files through makeinfo..."
-	make -C $(BUILDDIR)/texinfo info
-	@echo "makeinfo finished; the Info files are in $(BUILDDIR)/texinfo."
-
-gettext:
-	$(SPHINXBUILD) -b gettext $(I18NSPHINXOPTS) $(BUILDDIR)/locale
-	@echo
-	@echo "Build finished. The message catalogs are in $(BUILDDIR)/locale."
-
-changes:
-	$(SPHINXBUILD) -b changes $(ALLSPHINXOPTS) $(BUILDDIR)/changes
-	@echo
-	@echo "The overview file is in $(BUILDDIR)/changes."
-
-linkcheck:
-	$(SPHINXBUILD) -b linkcheck $(ALLSPHINXOPTS) $(BUILDDIR)/linkcheck
-	@echo
-	@echo "Link check complete; look for any errors in the above output " \
-	      "or in $(BUILDDIR)/linkcheck/output.txt."
-
-doctest:
-	$(SPHINXBUILD) -b doctest $(ALLSPHINXOPTS) $(BUILDDIR)/doctest
-	@echo "Testing of doctests in the sources finished, look at the " \
-	      "results in $(BUILDDIR)/doctest/output.txt."
-
-coverage:
-	$(SPHINXBUILD) -b coverage $(ALLSPHINXOPTS) $(BUILDDIR)/coverage
-	@echo "Testing of coverage in the sources finished, look at the " \
-	      "results in $(BUILDDIR)/coverage/python.txt."
-
-xml:
-	$(SPHINXBUILD) -b xml $(ALLSPHINXOPTS) $(BUILDDIR)/xml
-	@echo
-	@echo "Build finished. The XML files are in $(BUILDDIR)/xml."
-
-pseudoxml:
-	$(SPHINXBUILD) -b pseudoxml $(ALLSPHINXOPTS) $(BUILDDIR)/pseudoxml
-	@echo
-	@echo "Build finished. The pseudo-XML files are in $(BUILDDIR)/pseudoxml."
+	@echo "pdflatex finished; the PDF files are in $(BUILDDIR)-latex."

docs/api/python/conf.py

@@ -8,7 +8,6 @@
 #    <https://opensource.org/licenses/BSD-2-Clause>
 #
 # ------------------------------------------------------------
-
 """Sphinx configuration for Python API."""
 
 from __future__ import absolute_import
@@ -43,7 +42,9 @@ napoleon_use_rtype = False
 # Project
 project = 'dragon'
 copyright = 'Copyright (c) 2017-present, SeetaTech, Co.,Ltd'
-author = 'Ting Pan\\\\tingpan@seetatech.com'
+author = 'SeetaTech'
+with open('../../../dragon/version.txt', 'r') as f:
+    version = f.read().strip()
 
 # HTML
 html_theme = 'seeta'
@@ -60,17 +61,18 @@ html_theme_options = {
     'navbar_links': {
         'Install': path_to('../../install', 1),
         'API': [
-            ('C++', path_to('../cc', 1)),
-            ('Python', path_to('', 1))
+            ('master', path_to('../../api/python', 1)),
+            ('versions...', path_to('../../versions', 1)),
         ],
         'Github': 'https://github.com/seetaresearch/dragon',
     },
     'navbar_logo_link': path_to('../..', 1),
-    'sidebar_title': 'Python v0.3.0',
+    'sidebar_title': 'Python v{}'.format(version),
     'sidebar_title_link': path_to('../../versions', 1),
     'breadcrumb_links': [
         ('Dragon', path_to('../..', 1)),
         ('API', path_to('../../versions', 1)),
+        ('Dragon v{}'.format(version.replace('a0', '-a0')), path_to('../../api', 1)),
         ('Python', path_to('', 1)),
     ],
 }

docs/api/python/dragon/EagerTensor.rst

@@ -24,7 +24,7 @@ name
 
 ndim
 ####
-.. autoattribute:: dragon.EagerTensor.name
+.. autoattribute:: dragon.EagerTensor.ndim
 
 shape
 #####

docs/api/python/make.bat

-@ECHO OFF
+:: #########################################################
+:: Command file to build on Windows for Sphinx documentation
+:: #########################################################
 
-REM Command file for Sphinx documentation
+@echo off
 
+:: You can set these variables from the command line
 if "%SPHINXBUILD%" == "" (
 	set SPHINXBUILD=sphinx-build
 )
-set BUILDDIR=..\..\_build\api
-set ALLSPHINXOPTS=-d %BUILDDIR%/doctrees %SPHINXOPTS% .
-set I18NSPHINXOPTS=%SPHINXOPTS% .
+set BUILDDIR=..\..\_build\api\python
+set ALLSPHINXOPTS=-d %BUILDDIR%\doctrees %SPHINXOPTS% .
 if NOT "%PAPER%" == "" (
 	set ALLSPHINXOPTS=-D latex_paper_size=%PAPER% %ALLSPHINXOPTS%
-	set I18NSPHINXOPTS=-D latex_paper_size=%PAPER% %I18NSPHINXOPTS%
 )
 
 if "%1" == "" goto help
@@ -19,25 +20,9 @@ if "%1" == "help" (
 	:help
 	echo.Please use `make ^<target^>` where ^<target^> is one of
 	echo.  html       to make standalone HTML files
-	echo.  dirhtml    to make HTML files named index.html in directories
-	echo.  singlehtml to make a single large HTML file
-	echo.  pickle     to make pickle files
-	echo.  json       to make JSON files
-	echo.  htmlhelp   to make HTML files and a HTML help project
-	echo.  qthelp     to make HTML files and a qthelp project
-	echo.  devhelp    to make HTML files and a Devhelp project
-	echo.  epub       to make an epub
+	echo.  debughtml  to make debugging HTML files
 	echo.  latex      to make LaTeX files, you can set PAPER=a4 or PAPER=letter
-	echo.  text       to make text files
-	echo.  man        to make manual pages
-	echo.  texinfo    to make Texinfo files
-	echo.  gettext    to make PO message catalogs
-	echo.  changes    to make an overview over all changed/added/deprecated items
-	echo.  xml        to make Docutils-native XML files
-	echo.  pseudoxml  to make pseudoxml-XML files for display purposes
-	echo.  linkcheck  to check all external links for integrity
-	echo.  doctest    to run all doctests embedded in the documentation if enabled
-	echo.  coverage   to run coverage check of the documentation if enabled
+	echo.  latexpdf   to make LaTeX files and run them through pdflatex
 	goto end
 )
 
@@ -47,13 +32,7 @@ if "%1" == "clean" (
 	goto end
 )
 
-if "%2" == "f" (
-	for /d %%i in (%BUILDDIR%\*) do rmdir /q /s %%i
-	del /q /s %BUILDDIR%\*
-)
-
-
-REM Check if sphinx-build is available and fallback to Python version if any
+:: Check if sphinx-build is available and fallback to Python version if any
 %SPHINXBUILD% 2> nul
 if errorlevel 9009 goto sphinx_python
 goto sphinx_ok
@@ -76,192 +55,37 @@ if errorlevel 9009 (
 
 :sphinx_ok
 
-
 if "%1" == "html" (
-	%SPHINXBUILD% -b html %ALLSPHINXOPTS% %BUILDDIR%/python
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The HTML pages are in %BUILDDIR%/python.
-	goto end
-)
-
-if "%1" == "dirhtml" (
-	%SPHINXBUILD% -b dirhtml %ALLSPHINXOPTS% %BUILDDIR%/dirhtml
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The HTML pages are in %BUILDDIR%/dirhtml.
-	goto end
-)
-
-if "%1" == "singlehtml" (
-	%SPHINXBUILD% -b singlehtml %ALLSPHINXOPTS% %BUILDDIR%/singlehtml
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The HTML pages are in %BUILDDIR%/singlehtml.
-	goto end
-)
-
-if "%1" == "pickle" (
-	%SPHINXBUILD% -b pickle %ALLSPHINXOPTS% %BUILDDIR%/pickle
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished; now you can process the pickle files.
-	goto end
-)
-
-if "%1" == "json" (
-	%SPHINXBUILD% -b json %ALLSPHINXOPTS% %BUILDDIR%/json
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished; now you can process the JSON files.
-	goto end
-)
-
-if "%1" == "htmlhelp" (
-	%SPHINXBUILD% -b htmlhelp %ALLSPHINXOPTS% %BUILDDIR%/htmlhelp
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished; now you can run HTML Help Workshop with the ^
-.hhp project file in %BUILDDIR%/htmlhelp.
-	goto end
-)
-
-if "%1" == "qthelp" (
-	%SPHINXBUILD% -b qthelp %ALLSPHINXOPTS% %BUILDDIR%/qthelp
+	%SPHINXBUILD% -b html %ALLSPHINXOPTS% %BUILDDIR%
 	if errorlevel 1 exit /b 1
 	echo.
-	echo.Build finished; now you can run "qcollectiongenerator" with the ^
-.qhcp project file in %BUILDDIR%/qthelp, like this:
-	echo.^> qcollectiongenerator %BUILDDIR%\qthelp\Dragon.qhcp
-	echo.To view the help file:
-	echo.^> assistant -collectionFile %BUILDDIR%\qthelp\Dragon.ghc
+	echo.Build finished. The HTML pages are in %BUILDDIR%.
 	goto end
 )
 
-if "%1" == "devhelp" (
-	%SPHINXBUILD% -b devhelp %ALLSPHINXOPTS% %BUILDDIR%/devhelp
+if "%1" == "debughtml" (
+	%SPHINXBUILD% -b html -j %NUMBER_OF_PROCESSORS% %ALLSPHINXOPTS% %BUILDDIR%
 	if errorlevel 1 exit /b 1
 	echo.
-	echo.Build finished.
-	goto end
-)
-
-if "%1" == "epub" (
-	%SPHINXBUILD% -b epub %ALLSPHINXOPTS% %BUILDDIR%/epub
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The epub file is in %BUILDDIR%/epub.
+	echo.Build finished. The HTML pages are in %BUILDDIR%.
 	goto end
 )
 
 if "%1" == "latex" (
-	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%/latex
+	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%-latex
 	if errorlevel 1 exit /b 1
 	echo.
-	echo.Build finished; the LaTeX files are in %BUILDDIR%/latex.
+	echo.Build finished; the LaTeX files are in %BUILDDIR%-latex.
 	goto end
 )
 
 if "%1" == "latexpdf" (
-	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%/latex
-	cd %BUILDDIR%/latex
+	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%-latex
+	cd %BUILDDIR%-latex
 	make all-pdf
 	cd %~dp0
 	echo.
-	echo.Build finished; the PDF files are in %BUILDDIR%/latex.
-	goto end
-)
-
-if "%1" == "latexpdfja" (
-	%SPHINXBUILD% -b latex %ALLSPHINXOPTS% %BUILDDIR%/latex
-	cd %BUILDDIR%/latex
-	make all-pdf-ja
-	cd %~dp0
-	echo.
-	echo.Build finished; the PDF files are in %BUILDDIR%/latex.
-	goto end
-)
-
-if "%1" == "text" (
-	%SPHINXBUILD% -b text %ALLSPHINXOPTS% %BUILDDIR%/text
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The text files are in %BUILDDIR%/text.
-	goto end
-)
-
-if "%1" == "man" (
-	%SPHINXBUILD% -b man %ALLSPHINXOPTS% %BUILDDIR%/man
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The manual pages are in %BUILDDIR%/man.
-	goto end
-)
-
-if "%1" == "texinfo" (
-	%SPHINXBUILD% -b texinfo %ALLSPHINXOPTS% %BUILDDIR%/texinfo
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The Texinfo files are in %BUILDDIR%/texinfo.
-	goto end
-)
-
-if "%1" == "gettext" (
-	%SPHINXBUILD% -b gettext %I18NSPHINXOPTS% %BUILDDIR%/locale
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The message catalogs are in %BUILDDIR%/locale.
-	goto end
-)
-
-if "%1" == "changes" (
-	%SPHINXBUILD% -b changes %ALLSPHINXOPTS% %BUILDDIR%/changes
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.The overview file is in %BUILDDIR%/changes.
-	goto end
-)
-
-if "%1" == "linkcheck" (
-	%SPHINXBUILD% -b linkcheck %ALLSPHINXOPTS% %BUILDDIR%/linkcheck
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Link check complete; look for any errors in the above output ^
-or in %BUILDDIR%/linkcheck/output.txt.
-	goto end
-)
-
-if "%1" == "doctest" (
-	%SPHINXBUILD% -b doctest %ALLSPHINXOPTS% %BUILDDIR%/doctest
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Testing of doctests in the sources finished, look at the ^
-results in %BUILDDIR%/doctest/output.txt.
-	goto end
-)
-
-if "%1" == "coverage" (
-	%SPHINXBUILD% -b coverage %ALLSPHINXOPTS% %BUILDDIR%/coverage
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Testing of coverage in the sources finished, look at the ^
-results in %BUILDDIR%/coverage/python.txt.
-	goto end
-)
-
-if "%1" == "xml" (
-	%SPHINXBUILD% -b xml %ALLSPHINXOPTS% %BUILDDIR%/xml
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The XML files are in %BUILDDIR%/xml.
-	goto end
-)
-
-if "%1" == "pseudoxml" (
-	%SPHINXBUILD% -b pseudoxml %ALLSPHINXOPTS% %BUILDDIR%/pseudoxml
-	if errorlevel 1 exit /b 1
-	echo.
-	echo.Build finished. The pseudo-XML files are in %BUILDDIR%/pseudoxml.
+	echo.Build finished; the PDF files are in %BUILDDIR%-latex.
 	goto end
 )
 

docs/api/python/tensorflow/keras/regularizers.rst

@@ -12,7 +12,7 @@ regularizers
   `class L1L2 <regularizers/L1L2.html>`_
   : The L1L2 regularizer.
 
-  `class L2 <regularizers/L1.html>`_
+  `class L2 <regularizers/L2.html>`_
   : The L1 regularizer.
 
   `class Regularizer <regularizers/Regularizer.html>`_

dragon/core/context.h

@@ -17,15 +17,18 @@
 
 namespace dragon {
 
+/*!
+ * \brief The cpu device context.
+ */
 class DRAGON_API CPUContext {
  public:
   /*! \brief Default Constructor */
-  explicit CPUContext() : random_seed_(3) {}
+  CPUContext() : random_seed_(3) {}
 
-  /*! \brief Constructor with the specified random seed */
+  /*! \brief Constructor with the random seed */
   explicit CPUContext(unsigned int random_seed) : random_seed_(random_seed) {}
 
-  /*! \brief Constructor with the specified device option */
+  /*! \brief Constructor with the device option */
   explicit CPUContext(const DeviceOption& option)
       : random_seed_(
             option.has_random_seed() ? option.random_seed()
@@ -34,74 +37,74 @@ class DRAGON_API CPUContext {
   /*! \brief Destructor */
   virtual ~CPUContext() {}
 
-  /*! \brief Alloc the memory */
-  static void* New(size_t nbytes) {
-    void* data = malloc(nbytes);
-    CHECK(data) << "\nAllocate memory with " << nbytes << " bytes failed.";
+  /*! \brief Allocate a block of memory */
+  static void* New(size_t size) {
+    void* data = malloc(size);
+    CHECK(data) << "\nAllocate memory with " << size << " bytes failed.";
     return data;
   }
 
-  /*! \brief Zero-Reset the memory */
-  static void Memset(size_t nbytes, void* ptr) {
-    memset(ptr, 0, nbytes);
+  /*! \brief Set a memory block to the given value */
+  static void Memset(size_t n, void* ptr, int value = 0) {
+    memset(ptr, value, n);
   }
 
-  /*! \brief Copy the memory */
-  template <class DestContext, class SrcContext>
-  static void Memcpy(size_t nbytes, void* dest, const void* src) {
-    memcpy(dest, src, nbytes);
+  /*! \brief Set a memory block to the given value asynchronously */
+  void MemsetAsync(size_t n, void* ptr, int value) {
+    memset(ptr, value, n);
   }
 
-  /*! \brief Free the memory */
-  static void Delete(void* data) {
-    free(data);
+  /*! \brief Copy a memory block to the destination */
+  template <class DestContext, class SrcContext>
+  static void Memcpy(size_t n, void* dest, const void* src) {
+    memcpy(dest, src, n);
   }
 
-  /*! \brief Zero-Reset the memory asynchronously */
-  void MemsetAsync(size_t nbytes, void* ptr) {
-    memset(ptr, 0, nbytes);
+  /*! \brief Copy a memory block to the destination asynchronously */
+  template <class DestContext, class SrcContext>
+  void MemcpyAsync(size_t n, void* dest, const void* src) {
+    memcpy(dest, src, n);
   }
 
-  /*! \brief Copy the memory asynchronously */
-  template <class DestContext, class SrcContext>
-  void MemcpyAsync(size_t nbytes, void* dest, const void* src) {
-    memcpy(dest, src, nbytes);
+  /*! \brief Deallocate a memory block */
+  static void Delete(void* ptr) {
+    free(ptr);
   }
 
-  /*! \brief Switch to the device of this context */
+  /*! \brief Switch to the device in current thread */
   void SwitchToDevice() {}
 
-  /*! \brief Switch to the device with the given stream */
-  void SwitchToDevice(const int stream_id) {}
+  /*! \brief Switch to the device and select given stream in current thread */
+  void SwitchToDevice(int stream) {}
 
-  /*! \brief Copy the memory with given type asynchronously */
+  /*! \brief Copy a typed memory block to the destination */
   template <typename T, class DestContext, class SrcContext>
-  void Copy(int n, T* dest, const T* src) {
+  static void Copy(int n, T* dest, const T* src) {
     if (dest == src) return;
     if (std::is_fundamental<T>::value) {
       Memcpy<DestContext, SrcContext>(
           n * sizeof(T), (void*)dest, (const void*)src);
     } else {
-      for (int i = 0; i < n; i++) {
+      for (int i = 0; i < n; ++i) {
         dest[i] = src[i];
       }
     }
   }
 
-  /*! \brief Synchronize the dispatched operations */
+  /*! \brief Wait for the dispatched computation to complete */
   void FinishDeviceComputation() {}
 
   /*! \brief Return the device index */
-  int device_id() const {
+  int device() const {
     return 0;
   }
 
   /*! \brief Return the stream index */
-  int stream_id() const {
+  int stream() const {
     return 0;
   }
 
-  /*! \brief Return the internal random generator */
+  /*! \brief Return the random generator */
   std::mt19937* rand_generator() {
     if (!rand_generator_.get()) {
       rand_generator_.reset(new std::mt19937(random_seed_));
@@ -110,13 +113,13 @@ class DRAGON_API CPUContext {
   }
 
   /*! \brief Set the stream index */
-  void set_stream_id(int stream_id) {}
+  void set_stream(int stream) {}
 
  private:
-  /*! \brief Store the random seed */
+  /*! \brief The random seed */
   unsigned int random_seed_;
 
-  /*! \brief Store the internal random generator */
+  /*! \brief The random generator */
   unique_ptr<std::mt19937> rand_generator_;
 };
 

dragon/core/context_cnml.h

@@ -13,8 +13,6 @@
 #ifndef DRAGON_CORE_CONTEXT_CNML_H_
 #define DRAGON_CORE_CONTEXT_CNML_H_
 
-/* CAMBRICON CNRT && CNML Environment */
-
 #include "dragon/core/common.h"
 
 struct cnrtStream;
@@ -28,11 +26,19 @@ typedef struct cnmlFusionOp* cnmlFusionOp_t;
 
 namespace dragon {
 
-class CNRTObject;
-
+/*!
+ * \brief The cnml device context.
+ */
 class CNMLContext {
  public:
-  /*! \brief Default Constructor */
+  /*! \brief Default constructor */
+  CNMLContext() : device_id_(0), random_seed_(DEFAULT_RNG_SEED) {}
+
+  /*! \brief Constructor with the device index */
+  explicit CNMLContext(int device)
+      : device_id_(device), random_seed_(DEFAULT_RNG_SEED) {}
+
+  /*! \brief Constructor with the device option */
   explicit CNMLContext(const DeviceOption& option)
       : device_id_(option.device_id()),
         random_seed_(
@@ -41,77 +47,63 @@ class CNMLContext {
     CHECK_EQ(option.device_type(), PROTO_CNML);
   }
 
-  /*! \brief Constructor with the specified device index */
-  explicit CNMLContext(int device_id = 0)
-      : device_id_(device_id), random_seed_(DEFAULT_RNG_SEED) {}
+  /*! \brief Allocate a block of memory */
+  static void* New(size_t size) {
+    return nullptr;
+  }
 
-  /*! \brief Alloc the memory */
-  static void* New(size_t nbytes);
+  /*! \brief Set a memory block to the given value */
+  static void Memset(size_t n, void* ptr, int value) {}
 
-  /*! \brief Zero-Reset the memory */
-  static void Memset(size_t nbytes, void* ptr);
+  /*! \brief Set a memory block to the given value asynchronously */
+  void MemsetAsync(size_t n, void* ptr, int value) {
+    Memset(n, ptr, value);
+  }
 
-  /*! \brief Copy the memory */
+  /*! \brief Copy a memory block to the destination */
   template <class DestContext, class SrcContext>
-  static void Memcpy(size_t nbytes, void* dest, const void* src);
-
-  /*! \brief Free the memory */
-  static void Delete(void* data);
-
-  /*! \brief Zero-Reset the memory asynchronously */
-  void MemsetAsync(size_t nbytes, void* ptr) {
-    Memset(nbytes, ptr);
-  }
+  static void Memcpy(size_t n, void* dest, const void* src) {}
 
-  /*! \brief Copy the memory asynchronously */
+  /*! \brief Copy a memory block to the destination asynchronously */
   template <class DestContext, class SrcContext>
-  void MemcpyAsync(size_t nbytes, void* dest, const void* src) {
-    Memcpy<DestContext, SrcContext>(dest, src, nbytes);
+  void MemcpyAsync(size_t n, void* dest, const void* src) {
+    Memcpy<DestContext, SrcContext>(dest, src, n);
   }
 
-  /*! \brief Switch to the device with the given stream */
-  void SwitchToDevice(int stream_id) {}
+  /*! \brief Deallocate a memory block */
+  static void Delete(void* ptr) {}
 
-  /*! \brief Switch to the device of this context */
+  /*! \brief Switch to the device in current thread */
   void SwitchToDevice() {
     SwitchToDevice(0);
   }
 
-  /*! \brief Synchronize the dispatched operations */
-  void FinishDeviceComputation() {}
+  /*! \brief Switch to the device and select given stream in current thread */
+  void SwitchToDevice(int stream) {}
 
-  /*! \brief Return the specified cnrt stream */
-  static cnrtStream_t cnrt_stream(int device_id, int stream_id);
+  /*! \brief Wait for the dispatched computation to complete */
+  void FinishDeviceComputation() {}
 
-  /*! \brief Return the internal cnrt stream */
+  /*! \brief Return the cnrt stream */
   cnrtStream_t cnrt_stream() {
     return cnrt_stream(device_id_, stream_id_);
   }
 
+  /*! \brief Return the specified cnrt stream */
+  static cnrtStream_t cnrt_stream(int device_id, int stream_id) {
+    return (cnrtStream_t) nullptr;
+  }
+
   /*! \brief Return the device index */
-  int device_id() const {
+  int device() const {
     return device_id_;
   }
 
   /*! \brief Return the stream index */
-  int stream_id() const {
+  int stream() const {
     return stream_id_;
   }
 
-  /*! \brief Return the global context locker */
-  static std::mutex& mutex() {
-    static std::mutex m;
-    return m;
-  }
-
-  /*! \brief Return the thread local cnrt object */
-  static CNRTObject* obj();
-
-  /*! \brief Set the stream index */
-  void set_stream_id(int stream_id) {
-    stream_id_ = stream_id;
-  }
-
  private:
   int device_id_, stream_id_ = 1, random_seed_;
   unique_ptr<std::mt19937> rand_generator_;

dragon/core/graph.cc

+#include <regex>
+
 #include "dragon/core/graph.h"
 #include "dragon/core/graph_gradient.h"
 #include "dragon/core/graph_optimizer.h"
@@ -46,8 +48,8 @@ GraphBase::GraphBase(const GraphDef& def, Workspace* ws)
   }
 }
 
-bool Graph::Create(const GraphDef& def, Workspace* ws) {
-  this->opt_def_ = def; // Store for debugging
+bool Graph::Create(const GraphDef& def) {
+  this->optimized_def_ = def; // Store for debugging
   bool has_device_option = def.has_device_option();
   for (int i = 0; i < def.op_size(); i++) {
     auto op_def(def.op(i));
@@ -63,7 +65,7 @@ bool Graph::Create(const GraphDef& def, Workspace* ws) {
       arg.set_i(1);
       op_def.add_arg()->CopyFrom(arg);
     }
-    cached_ops_.push_back(NewOperator(op_def, ws));
+    cached_ops_.push_back(NewOperator(op_def, ws_));
     cached_ops_.back()->set_output_aliases(output_aliases_);
   }
   return true;
@@ -71,25 +73,25 @@ bool Graph::Create(const GraphDef& def, Workspace* ws) {
 
 Graph::Graph(const GraphDef& def, Workspace* ws) : GraphBase(def, ws) {
   // Apply the optimizations
-  GraphDef opt_def = def;
-  GraphOptimizer graph_optim(ws);
+  GraphDef def_v2(def);
+  GraphOptimizer graph_optimizer(ws);
   GraphGradientMaker gradient_maker;
   Map<string, vec32_t> subgraph_indices;
-  int opt = 3; // defaults: O3
+  int opt = 3; // default: O3
   if (args().count("optimization")) opt = arg("optimization").i();
-  if (opt >= 1) opt_def = graph_optim.PruneNodes(def);
-  if (opt >= 2) graph_optim.AddInplace(opt_def, output_aliases_);
+  if (opt >= 1) def_v2 = graph_optimizer.PruneNodes(def);
+  if (opt >= 2) graph_optimizer.AddInplace(def_v2, output_aliases_);
   if (opt >= 3) {
     if (phase() == "TRAIN") {
-      opt_def = graph_optim.MirrorStage(opt_def, subgraph_indices);
-      opt_def = gradient_maker.Share(opt_def);
+      def_v2 = graph_optimizer.MirrorStage(def_v2, subgraph_indices);
+      def_v2 = gradient_maker.Share(def_v2);
     } else {
-      opt_def = graph_optim.SimulateGC(opt_def);
+      def_v2 = graph_optimizer.SimulateGC(def_v2);
     }
   }
 
   // Create
-  Create(opt_def, ws);
+  Create(def_v2);
 
   // Recomputation and SubGraph
   if (subgraph_indices.size() > 0) {
@@ -105,11 +107,14 @@ Graph::Graph(const GraphDef& def, Workspace* ws) : GraphBase(def, ws) {
   }
 }
 
-bool Graph::Run(const string& include, const string& exclude, int stream) {
+bool Graph::Run(int stream, const string& include, const string& exclude) {
+  unique_ptr<std::regex> regex_incl, regex_excl;
+  if (!include.empty()) regex_incl.reset(new std::regex(include));
+  if (!exclude.empty()) regex_excl.reset(new std::regex(exclude));
   LOG(DEBUG) << "Run Graph: " << name();
   for (auto* op : cached_ops_) {
-    if (!include.empty() && !str::find(op->type(), include)) continue;
-    if (!exclude.empty() && str::find(op->type(), exclude)) continue;
+    if (regex_incl && !regex_match(op->type(), *regex_incl)) continue;
+    if (regex_excl && regex_match(op->type(), *regex_excl)) continue;
     op->SwitchToPhase(phase());
     LOG(DEBUG) << "Run Op: " << op->name();
     op->Run(stream);

dragon/core/graph.h

@@ -18,26 +18,32 @@
 
 namespace dragon {
 
+/*!
+ * \brief The base graph class.
+ */
 class DRAGON_API GraphBase {
  public:
-  /*! \brief Default constructor */
-  GraphBase(const GraphDef&, Workspace*);
+  /*! \brief Constructor with the def and workspace */
+  GraphBase(const GraphDef& def, Workspace* ws);
 
-  /*! \brief Default Destructor */
+  /*! \brief Destructor */
   virtual ~GraphBase() {}
 
-  /*! \brief Create a graph from the optimized def */
-  virtual bool Create(const GraphDef&, Workspace*) = 0;
+  /*! \brief Create graph in the workspace */
+  virtual bool Create(const GraphDef& def) = 0;
 
-  /*! \brief Run the graph once synchronously */
-  virtual bool Run(const string&, const string&, int = 0) = 0;
+  /*! \brief Run graph on the given stream */
+  virtual bool Run(
+      int stream = 0,
+      const string& include = "",
+      const string& exclude = "") = 0;
 
   /*! \brief Return the graph name */
   const string& name() const {
     return name_;
   }
 
-  /*! \brief Return the defined running phase */
+  /*! \brief Return the executing phase */
   const string& phase() const {
     return phase_;
   }
@@ -47,19 +53,19 @@ class DRAGON_API GraphBase {
     return *(args_[name]);
   }
 
-  /*! \brief Return the argument map */
+  /*! \brief Return all the arguments */
   const Map<string, const Argument*>& args() {
     return args_;
   }
 
-  /*! \brief Return the stored raw def */
+  /*! \brief Return the graph def */
   const GraphDef& def() const {
     return def_;
   }
 
-  /*! \brief Return the stored opt def */
-  const GraphDef& opt_def() const {
-    return opt_def_;
+  /*! \brief Return the optimized graph def */
+  const GraphDef& optimized_def() const {
+    return optimized_def_;
   }
 
   /*! \brief Return the parent workspace */
@@ -68,42 +74,53 @@ class DRAGON_API GraphBase {
   }
 
  protected:
-  /*! \brief Store the name and running phase */
+  /*! \brief The name and executing phase */
   string name_, phase_;
 
-  /*! \brief Store the defined arguments */
+  /*! \brief The defined arguments */
   Map<string, const Argument*> args_;
 
-  /*! \brief Store the parent workspace */
+  /*! \brief The parent workspace */
   Workspace* ws_;
 
-  /*! \brief Store the graph definition */
-  GraphDef def_, opt_def_;
+  /*! \brief The graph definition */
+  GraphDef def_;
+
+  /*! \brief The optimized graph definition */
+  GraphDef optimized_def_;
+
+  DISABLE_COPY_AND_ASSIGN(GraphBase);
 };
 
+/*!
+ * \brief Graph to execute operators sequentially.
+ */
 class Graph : public GraphBase {
  public:
-  /*! \brief Default constructor */
+  /*! \brief Constructor with the def and workspace */
   Graph(const GraphDef& def, Workspace* ws);
 
-  /*! \brief Default Destructor */
+  /*! \brief Destructor */
   virtual ~Graph() {
     for (auto* cached_op : cached_ops_) {
       delete cached_op;
     }
   }
 
-  /*! \brief Create a graph from the optimized def */
-  bool Create(const GraphDef&, Workspace*) override;
+  /*! \brief Create graph in the workspace */
+  bool Create(const GraphDef& def) override;
 
-  /*! \brief Run the graph once synchronously */
-  bool Run(const string&, const string&, int = 0) override;
+  /*! \brief Run graph on the given stream */
+  bool Run(
+      int stream = 0,
+      const string& include = "",
+      const string& exclude = "") override;
 
  protected:
   /*! \brief The cached operators */
   vector<OperatorBase*> cached_ops_;
 
-  /*! \brief Store the candidate output aliases */
+  /*! \brief The candidate output aliases */
   Map<string, Set<string>> output_aliases_;
 };
 

dragon/core/graph_optimizer.h

@@ -72,6 +72,9 @@ class GraphOptimizer {
 
   /* \brief Store the count of references */
   Map<string, int> reference_count_;
+
+ private:
+  DISABLE_COPY_AND_ASSIGN(GraphOptimizer);
 };
 
 } // namespace dragon

dragon/core/memory.h

@@ -20,44 +20,45 @@
 namespace dragon {
 
 typedef enum {
-  NCHW,
-  NHWC,
+  NCHW = 0,
+  NHWC = 1,
 } StorageOrder;
 
+/*!
+ * \brief Memory to manage both the host and device data.
+ */
 class DRAGON_API UnifiedMemory {
  public:
-  typedef enum {
-    /*! \brief The initial state */
-    UNINITIALIZED,
-    /*! \brief Memory could be modified by CPUContext last time */
-    STATE_AT_CPU,
-    /*! \brief Memory could be modified by CUDAContext last time */
-    STATE_AT_CUDA,
-    /*! \brief Memory could be modified by CNMLContext last time */
-    STATE_AT_CNML,
-    /*! \brief The synced state  */
-    SYNCED,
-  } State;
-
-  /*! \brief Default Constructor */
-  UnifiedMemory() : cpu_ptr_(nullptr), cuda_ptr_(nullptr), cnml_ptr_(nullptr) {}
-
-  /*! \brief Constructor with the known meta and size */
-  UnifiedMemory(const TypeMeta& meta, size_t size)
-      : meta_(meta),
-        size_(size),
-        cpu_ptr_(nullptr),
-        cuda_ptr_(nullptr),
-        cnml_ptr_(nullptr) {}
+  /*!
+   * \brief The device-aware state for data mutation.
+   */
+  enum State {
+    /*! \brief Initial state */
+    UNINITIALIZED = 0,
+    /*! \brief Data is mutable to cpu */
+    STATE_AT_CPU = 1,
+    /*! \brief Data is mutable to cuda */
+    STATE_AT_CUDA = 2,
+    /*! \brief Data is mutable to cnml */
+    STATE_AT_CNML = 3,
+    /*! \brief Data is synced between host and device */
+    SYNCED = 4,
+  };
+
+  /*! \brief Default constructor */
+  UnifiedMemory() {}
+
+  /*! \brief Constructor with the type meta and size */
+  UnifiedMemory(const TypeMeta& meta, size_t size) : meta_(meta), size_(size) {}
 
   /*! \brief Destructor */
   ~UnifiedMemory();
 
-  /*! \brief Switch to the specified device */
-  void SwitchToDevice(int device_id);
+  /*! \brief Switch to the given device */
+  void SwitchToDevice(int device);
 
-  /*! \brief Switch to the specified cuda device */
-  void SwitchToCUDADevice(int device_id);
+  /*! \brief Switch to the given cuda device */
+  void SwitchToCUDADevice(int device);
 
   /*! \brief Involve the state to CPUContext */
   void ToCPU(size_t size = 0);
@@ -65,9 +66,9 @@ class DRAGON_API UnifiedMemory {
   /*! \brief Involve the state to CUDAContext */
   void ToCUDA(size_t size = 0);
 
-  /*! \brief Return the device index */
-  int device_id() const {
-    return device_id_;
+  /*! \brief Return the memory state */
+  State state() const {
+    return state_;
   }
 
   /*! \brief Return the total number of bytes */
@@ -75,9 +76,9 @@ class DRAGON_API UnifiedMemory {
     return size_;
   }
 
-  /*! \brief Return the number of chunks */
-  size_t nchunks() const {
-    return nchunks_;
+  /*! \brief Return the number of memory chunks */
+  size_t num_chunks() const {
+    return num_chunks_;
   }
 
   /*! \brief Return the storage order */
@@ -85,30 +86,30 @@ class DRAGON_API UnifiedMemory {
     return order_;
   }
 
-  /*! \brief Return the memory state */
-  State state() const {
-    return state_;
+  /*! \brief Return the device index */
+  int device() const {
+    return device_id_;
   }
 
-  /*! \brief Return a string to describe the internal structure */
+  /*! \brief Return the data info */
   Map<string, string> info() const;
 
-  /*! \brief Return the const data pointer on CPUContext */
+  /*! \brief Return the const cpu data */
   const void* cpu_data(size_t size = 0);
 
-  /*! \brief Return the const data pointer on CUDAContext */
+  /*! \brief Return the const cuda data */
   const void* cuda_data(size_t size = 0);
 
-  /*! \brief Return the const data pointer on CNMLContext */
+  /*! \brief Return the const cnml data */
   const void* cnml_data();
 
-  /*! \brief Return the mutable data pointer on CPUContext */
+  /*! \brief Return the mutable cpu data */
   void* mutable_cpu_data(size_t size = 0);
 
-  /*! \brief Return the mutable data pointer on CUDAContext */
+  /*! \brief Return the mutable cuda data */
   void* mutable_cuda_data(size_t size = 0);
 
-  /*! \brief Return the mutable data pointer on CNMLContext */
+  /*! \brief Return the mutable cnml data */
   void* mutable_cnml_data();
 
   /*! \brief Return the binding cnml cpu tensor */
@@ -117,15 +118,15 @@ class DRAGON_API UnifiedMemory {
   /*! \brief Return the binding cnml mlu tensor */
   cnmlTensor_t& cnml_mlu_tensor();
 
-  /*! \brief Allocate the mlu device memory */
+  /*! \brief Allocate the mlu device data */
   void* malloc_cnml_data();
 
-  /*! \brief Copy the mlu device memory to the host */
+  /*! \brief Copy the mlu device data to host */
   void fetch_cnml_data(void** data);
 
-  /*! \brief Set the chunks of this memory */
-  void set_nchunks(size_t nchunks) {
-    nchunks_ = nchunks;
+  /*! \brief Set the number of data chunks */
+  void set_num_chunks(size_t num_chunks) {
+    num_chunks_ = num_chunks;
   }
 
   /*! \brief Set the storage order */
@@ -133,39 +134,47 @@ class DRAGON_API UnifiedMemory {
     order_ = order;
   }
 
-  /*! \brief Set the cpu data pointer from external context */
+  /*! \brief Set to use an external block of cpu data */
   void set_cpu_data(void* cpu_ptr, size_t size);
 
-  /*! \brief Set the cuda data pointer from external context */
-  void set_cuda_data(void* cuda_ptr, size_t size, int device_id);
+  /*! \brief Set to use an extenral block of cuda data */
+  void set_cuda_data(void* cuda_ptr, size_t size, int device);
 
  private:
-  /*! \brief The type meta */
-  TypeMeta meta_;
+  /*! \brief The data state */
+  State state_ = UNINITIALIZED;
 
   /*! \brief The size and number of chunks */
-  size_t size_ = 0, nchunks_ = 1;
+  size_t size_ = 0, num_chunks_ = 1;
+
+  /*! \brief The type meta */
+  TypeMeta meta_;
 
   /*! \brief The storage order */
   StorageOrder order_ = NCHW;
 
-  /*! \brief The current state */
-  State state_ = UNINITIALIZED;
+  /*! \brief The device index */
+  int device_id_ = 0;
+
+  /*! \brief The cpu data pointer */
+  void* cpu_ptr_ = nullptr;
 
-  /*! \brief The data pointers */
-  void *cpu_ptr_, *cuda_ptr_, *cnml_ptr_;
+  /*! \brief The cuda data pointer */
+  void* cuda_ptr_ = nullptr;
+
+  /*! \brief The cnml data pointer */
+  void* cnml_ptr_ = nullptr;
 
   /*! \brief The ownership of data pointers */
   int own_cpu_ptr_ = 1, own_cuda_ptr_ = 1;
 
-  /*! \brief The device index */
-  int device_id_ = 0;
-
   /*! \brief The binding cpu tensor for cnml */
   cnmlCpuTensor_t cnml_cpu_tensor_ = nullptr;
 
   /*! \brief The binding mlu tensor for cnml */
   cnmlTensor_t cnml_mlu_tensor_ = nullptr;
+
+  DISABLE_COPY_AND_ASSIGN(UnifiedMemory);
 };
 
 } // namespace dragon

dragon/core/operator.cc

@@ -41,12 +41,6 @@ OperatorBase::OperatorBase(const OperatorDef& def, Workspace* ws)
   }
 }
 
-// template <class Context>
-// Operator<Context>::Operator(const OperatorDef& def, Workspace* ws)
-//     : OperatorBase(def, ws),
-//       ctx_(def.device_option()),
-//       do_sync_(OpArg<bool>("do_sync", false)) {}
-
 Tensor& OperatorBase::Input(int i) {
   CHECK_LT(i, (int)inputs_.size());
   CHECK_GE(i, -(int)inputs_.size());
@@ -80,27 +74,17 @@ Tensor* OperatorBase::Buffer(const string& name) {
   return ws()->CreateTensor("/share/buffer/" + handle_ + "/" + name);
 }
 
-string OperatorBase::TypeString(const Tensor& tensor, const Set<string>& types)
-    const {
-  std::stringstream ss;
-  ss << "Unsupported type of Tensor(" << tensor.name()
-     << "): " << types::to_string(tensor.meta()) << "\n";
-  ss << "<" << type() << "Op>"
-     << " supports the following types: {\n";
-  for (auto& type : types)
-    ss << "  * " << type << ",\n";
-  ss << "}";
-  return ss.str();
-}
-
-string OperatorBase::TypeString(const string& dtype, const Set<string>& types)
-    const {
+string OperatorBase::MessageForUnsupported(
+    const string& value,
+    const vector<string>& support_values,
+    const string& entry) const {
   std::stringstream ss;
-  ss << "Unsupported type: " << dtype << "\n";
+  ss << "Unsupported " << entry << ": " << value << "\n";
   ss << "<" << type() << "Op>"
-     << " supports the following types: {\n";
-  for (auto& type : types)
-    ss << "  * " << type << ",\n";
+     << " supports the following " << entry << "(s): {\n";
+  for (const auto& support_value : support_values) {
+    ss << "  * " << support_value << ",\n";
+  }
   ss << "}";
   return ss.str();
 }
@@ -133,7 +117,7 @@ void Operator<Context>::Prepare() {
       flag->mutable_data<bool, CPUContext>()[0] = true;
       vector<OperatorBase*>& chain = subgraph()[name];
       for (auto* op : chain) {
-        op->Run(ctx()->stream_id());
+        op->Run(ctx()->stream());
       }
       flag->mutable_data<bool, CPUContext>()[0] = false;
     }
@@ -156,12 +140,12 @@ template <class Context>
 void Operator<Context>::SwitchToDevice() {
   for (auto* tensor : inputs_) {
     if (tensor->has_name()) {
-      tensor->SwitchToDevice(ctx()->device_id());
+      tensor->SwitchToDevice(ctx()->device());
     }
   }
   for (auto* tensor : outputs_) {
     if (tensor->has_name()) {
-      tensor->SwitchToDevice(ctx()->device_id());
+      tensor->SwitchToDevice(ctx()->device());
     }
   }
 }

dragon/core/operator.h

@@ -28,30 +28,30 @@ class DRAGON_API OperatorBase {
  public:
   typedef Map<string, vector<OperatorBase*>> SubGraph;
 
-  /*! \brief Default constructor */
+  /*! \brief Constructor with the def and workspace */
   OperatorBase(const OperatorDef&, Workspace*);
 
-  /*! \brief Default Destructor */
+  /*! \brief Destructor */
   virtual ~OperatorBase() {}
 
-  /*! \brief Fusion this operator into the specified graph */
-  virtual void Fusion(void* graph) {
+  /*! \brief Update operator from the given def  */
+  OperatorBase* UpdateFrom(const OperatorDef&);
+
+  /*! \brief Fusion operator into the given graph */
+  virtual void Fuse(void* graph) {
     NOT_IMPLEMENTED;
   }
 
-  /*! \brief Run operator on the specified stream */
+  /*! \brief Run operator on the given stream */
   virtual void Run(int stream = 0) {
     NOT_IMPLEMENTED;
   }
 
-  /*! \brief Switch the internal running phase */
+  /*! \brief Switch to the given executing phase */
   void SwitchToPhase(const string& phase) {
     phase_ = phase;
   }
 
-  /*! \brief Update operator according to a new def  */
-  OperatorBase* UpdateFrom(const OperatorDef&);
-
   /*! \brief Return the input tensor */
   Tensor& Input(int i);
 
@@ -61,7 +61,7 @@ class DRAGON_API OperatorBase {
   /*! \brief Return the output tensor with input aliases */
   Tensor* Output(int i, const vec32_t& inputs);
 
-  /*! \brief Return the unique named buffer */
+  /*! \brief Return the buffer tensor */
   Tensor* Buffer(const string& name);
 
   /*! \brief Return the number of inputs */
@@ -74,31 +74,26 @@ class DRAGON_API OperatorBase {
     return (int)outputs_.size();
   }
 
-  /*! \brief Return the value of the specified argument */
+  /*! \brief Return the value of single argument */
   template <typename T>
   T Arg(const string& name, const T& default_value);
 
-  /*! \brief Return the values of the specified argument */
+  /*! \brief Return the value of repeated argument */
   template <typename T>
   vector<T> Args(const string& name);
 
-  /*! \brief Return the debug string of stored def */
-  string DebugString() const {
-    return def_.DebugString();
-  }
-
-  /*! \brief Return the debug string of tensor type */
-  string TypeString(const Tensor&, const Set<string>&) const;
-
-  /* \brief Return the debug string of given type */
-  string TypeString(const string&, const Set<string>&) const;
+  /*! \brief Return the message for supported value */
+  string MessageForUnsupported(
+      const string& value,
+      const vector<string>& support_values,
+      const string& entry = "type") const;
 
   /*! \brief Return the specified argument */
   const Argument& arg(const string& name) {
     return *(args_[name]);
   }
 
-  /*! \brief Return the argument map */
+  /*! \brief Return all the arguments */
   const Map<string, const Argument*>& args() {
     return args_;
   }
@@ -113,7 +108,7 @@ class DRAGON_API OperatorBase {
     return def_.type();
   }
 
-  /*! \brief Return the current running phase */
+  /*! \brief Return the running phase */
   const string& phase() const {
     return phase_;
   }
@@ -190,12 +185,17 @@ class DRAGON_API OperatorBase {
 
   /*! \brief Store the defined arguments */
   Map<string, const Argument*> args_;
+
+  DISABLE_COPY_AND_ASSIGN(OperatorBase);
 };
 
+/*!
+ * \brief The base operator class with context.
+ */
 template <class Context>
 class DRAGON_API Operator : public OperatorBase {
  public:
-  /*! \brief Default constructor */
+  /*! \brief Constructor with the def and workspace */
   Operator(const OperatorDef& def, Workspace* ws)
       : OperatorBase(def, ws),
         ctx_(def.device_option()),
@@ -254,8 +254,7 @@ OperatorBase* NewOperator(const OperatorDef&, Workspace*);
   using OperatorBase::Buffer;                \
   using OperatorBase::InputSize;             \
   using OperatorBase::OutputSize;            \
-  using OperatorBase::DebugString;   \
-  using OperatorBase::TypeString;    \
+  using OperatorBase::MessageForUnsupported; \
   using OperatorBase::name;                  \
   using OperatorBase::type;                  \
   using OperatorBase::phase;                 \
@@ -323,10 +322,9 @@ struct DispatchHelper;
   struct DispatchHelper<TensorTypes<>, Args...> {                          \
     template <typename Op>                                                 \
     static void Call(Op* op, const TypeMeta& meta, string& types) {        \
-      LOG(FATAL) << "Unsupported tensor type: " << types::to_string(meta) \
-                 << "\n"                                                  \
+      LOG(FATAL) << "Unsupported type: " << types::to_string(meta) << "\n" \
                  << "<" << op->type() << "Op>"                             \
-                 << " supports the following types: {\n"                  \
+                 << " supports the following type(s): {\n"                 \
                  << types << "}";                                          \
     }                                                                      \
     template <typename Op>                                                 \

dragon/core/registry.h

@@ -17,78 +17,93 @@
 
 namespace dragon {
 
-template <class SrcType, class ObjType, class... Args>
+/*!
+ * \brief Registry to create class instances.
+ */
+template <class KeyType, class ObjectType, class... Args>
 class Registry {
  public:
-  typedef std::function<ObjType*(Args...)> Creator;
+  typedef std::function<ObjectType*(Args...)> Creator;
 
-  ObjType* Create(const SrcType& key, Args... args) {
+  /*! \brief Create an instance of specified class */
+  ObjectType* Create(const KeyType& key, Args... args) {
     CHECK(registry_.count(key)) << "\nKey(" << key << ") has not registered.";
     return registry_[key](args...);
   }
 
-  bool Has(const SrcType& key) {
+  /*! \brief Return whether the specified class is registered */
+  bool Has(const KeyType& key) {
     return (registry_.count(key)) != 0;
   }
 
-  void Register(const SrcType& key, Creator creator) {
+  /*! \brief Register a class with the creator */
+  void Register(const KeyType& key, Creator creator) {
     CHECK(!registry_.count(key))
         << "\nKey(" << key << ") has already registered.";
     registry_[key] = creator;
   }
 
-  vector<SrcType> keys() {
-    vector<SrcType> ret;
-    for (const auto& it : registry_)
+  /*! \brief Return the key of registered classes */
+  vector<KeyType> keys() {
+    vector<KeyType> ret;
+    for (const auto& it : registry_) {
       ret.push_back(it.first);
+    }
     return ret;
   }
 
  private:
-  Map<SrcType, Creator> registry_;
+  /*! \brief The registry map */
+  Map<KeyType, Creator> registry_;
 };
 
-template <class SrcType, class ObjType, class... Args>
+/*!
+ * \brief Register creator into the registry.
+ */
+template <class KeyType, class ObjectType, class... Args>
 class Registerer {
  public:
+  /*! \brief Constructor with key and creator */
   Registerer(
-      const SrcType& key,
-      Registry<SrcType, ObjType, Args...>* registry,
-      typename Registry<SrcType, ObjType, Args...>::Creator creator,
+      const KeyType& key,
+      Registry<KeyType, ObjectType, Args...>* registry,
+      typename Registry<KeyType, ObjectType, Args...>::Creator creator,
       const string& help_msg = "") {
     registry->Register(key, creator);
   }
 
+  /*! \brief Return the default creator */
   template <class DerivedType>
-  static ObjType* defaultCreator(Args... args) {
+  static ObjectType* DefaultCreator(Args... args) {
     return new DerivedType(args...);
   }
 };
 
 // Used in *.h files
-#define DECLARE_TYPED_REGISTRY(RegistryName, SrcType, ObjType, ...)     \
-  DRAGON_API Registry<SrcType, ObjType, ##__VA_ARGS__>* RegistryName(); \
-  typedef Registerer<SrcType, ObjType, ##__VA_ARGS__> Registerer##RegistryName;
+#define DECLARE_TYPED_REGISTRY(RegistryName, KeyType, ObjectType, ...)     \
+  DRAGON_API Registry<KeyType, ObjectType, ##__VA_ARGS__>* RegistryName(); \
+  typedef Registerer<KeyType, ObjectType, ##__VA_ARGS__>                   \
+      Registerer##RegistryName;
 
 // Used in *.cc files
-#define DEFINE_TYPED_REGISTRY(RegistryName, SrcType, ObjType, ...) \
-  Registry<SrcType, ObjType, ##__VA_ARGS__>* RegistryName() {      \
-    static Registry<SrcType, ObjType, ##__VA_ARGS__>* registry =   \
-        new Registry<SrcType, ObjType, ##__VA_ARGS__>();           \
+#define DEFINE_TYPED_REGISTRY(RegistryName, KeyType, ObjectType, ...) \
+  Registry<KeyType, ObjectType, ##__VA_ARGS__>* RegistryName() {      \
+    static Registry<KeyType, ObjectType, ##__VA_ARGS__>* registry =   \
+        new Registry<KeyType, ObjectType, ##__VA_ARGS__>();           \
     return registry;                                                  \
   }
 
-#define DECLARE_REGISTRY(RegistryName, ObjType, ...) \
-  DECLARE_TYPED_REGISTRY(RegistryName, string, ObjType, ##__VA_ARGS__)
+#define DECLARE_REGISTRY(RegistryName, ObjectType, ...) \
+  DECLARE_TYPED_REGISTRY(RegistryName, string, ObjectType, ##__VA_ARGS__)
 
-#define DEFINE_REGISTRY(RegistryName, ObjType, ...) \
-  DEFINE_TYPED_REGISTRY(RegistryName, string, ObjType, ##__VA_ARGS__)
+#define DEFINE_REGISTRY(RegistryName, ObjectType, ...) \
+  DEFINE_TYPED_REGISTRY(RegistryName, string, ObjectType, ##__VA_ARGS__)
 
 #define REGISTER_TYPED_CLASS(RegistryName, key, ...)                    \
   static Registerer##RegistryName ANONYMOUS_VARIABLE(g_##RegistryName)( \
       key,                                                              \
       RegistryName(),                                                   \
-      Registerer##RegistryName::defaultCreator<__VA_ARGS__>)
+      Registerer##RegistryName::DefaultCreator<__VA_ARGS__>)
 
 #define REGISTER_CLASS(RegistryName, key, ...) \
   REGISTER_TYPED_CLASS(RegistryName, #key, __VA_ARGS__)

dragon/core/tensor.h

@@ -18,28 +18,53 @@
 
 namespace dragon {
 
+/*!
+ * \brief The base tensor class, manage memory or not.
+ *
+ * Tensor is usually constructed with the shape info:
+ *
+ * \code{.cpp}
+ * auto* a = new dragon::Tensor(std::vector<int64_t>({2, 3}));
+ * auto* b = dragon::Tensor().Reshape({2, 3}); // Equivalent
+ * \endcode
+ *
+ * To allocate the data, type meta and device context are also required:
+ *
+ * \code{.cpp}
+ * auto meta = dragon::TypeMeta::Make<float>();
+ * auto* raw_data = a->raw_mutable_data<dragon::CPUContext>(meta);
+ * auto* data = b->mutable_data<float, dragon::CPUContext>();
+ * \endcode
+ *
+ * Memory will be reset if required bytes is larger than capacity:
+ * \code{.cpp}
+ * std::cout << a->nbytes() << " " << a->capacity() << std::endl; // 24, 24
+ * std::cout << a->Reshape({2, 4})->size() << std::endl; // 8
+ * std::cout << a->nbytes() << " " << a->capacity() << std::endl; // 32, 0
+ * a->mutable_data<float, dragon::CPUContext>();
+ * a->Reshape({2, 3});
+ * std::cout << a->nbytes() << " " << a->capacity() << std::endl; // 24, 32
+ * \endcode
+ */
 class DRAGON_API Tensor {
  public:
-  Tensor(const Tensor&) = delete;
-  Tensor& operator=(const Tensor&) = delete;
-
-  /*! \brief Default Constructor */
+  /*! \brief Default constructor */
   Tensor() : name_("") {}
 
-  /*! \brief Constructor with the known name */
+  /*! \brief Constructor with the name */
   explicit Tensor(const string& name) : name_(name) {}
 
-  /*! \brief Constructor with the known int64 dimensions */
+  /*! \brief Constructor with the int64 dimensions */
   explicit Tensor(const vec64_t& dims) {
     Reshape(dims);
   }
 
-  /*! \brief Constructor with the known int32 dimensions */
+  /*! \brief Constructor with the int32 dimensions */
   explicit Tensor(const vec32_t& dims) {
     Reshape(vec64_t(dims.begin(), dims.end()));
   }
 
-  /*! \brief Constructor with the known meta */
+  /*! \brief Constructor with the type meta */
   explicit Tensor(const TypeMeta& meta) {
     set_meta(meta);
   }
@@ -54,7 +79,7 @@ class DRAGON_API Tensor {
     }
   }
 
-  /*! \brief Reshape to the given dimensions */
+  /*! \brief Change the tensor dimensions */
   Tensor* Reshape(const vec64_t& dims) {
     dims_ = dims;
     strides_.resize(dims.size());
@@ -79,18 +104,18 @@ class DRAGON_API Tensor {
     return this;
   }
 
-  /*! \brief Reshape the dimensions like the given tensor */
+  /*! \brief Change the tensor dimensions as the other */
   Tensor* ReshapeLike(const Tensor& other) {
     return Reshape(other.dims_);
   }
 
-  /*! \brief Switch the memory to the specific device */
+  /*! \brief Switch memory to the specific device */
   void SwitchToDevice(int device_id) {
     UnifiedMemory* mem = memory();
     if (mem) mem->SwitchToDevice(device_id);
   }
 
-  /*! \brief Copy memory from the tensor with context */
+  /*! \brief Copy memory from a tensor with context */
   template <class Context>
   Tensor* CopyFrom(const Tensor& other, Context* ctx) {
     if ((void*)&other == (void*)this) return this;
@@ -102,7 +127,7 @@ class DRAGON_API Tensor {
     return this;
   }
 
-  /*! \brief Copy memory from the vector */
+  /*! \brief Copy memory from a vector */
   template <typename TensorType, typename VectorType>
   Tensor* CopyFrom(const vector<VectorType>& other) {
     if (other.size() > 0) {
@@ -115,7 +140,7 @@ class DRAGON_API Tensor {
     return this;
   }
 
-  /*! \brief Copy memory to the vector */
+  /*! \brief Copy memory to a vector */
   template <typename TensorType, typename VectorType>
   void CopyTo(vector<VectorType>& dest) {
     dest.resize(size());
@@ -141,7 +166,7 @@ class DRAGON_API Tensor {
     own_memory_ = (memory == nullptr);
   }
 
-  /*! \brief Reset all resources */
+  /*! \brief Reset tensor to release all resources */
   void Reset() {
     dims_.clear();
     strides_.clear();
@@ -156,13 +181,13 @@ class DRAGON_API Tensor {
     }
   }
 
-  /*! \brief Whether the data type is matched */
+  /*! \brief Return whether the data type is matched */
   template <typename T>
   bool IsType() {
     return meta_.Match<T>();
   }
 
-  /*! \brief Return a string formatting the dimensions */
+  /*! \brief Return a string formatting the given dimensions */
   static string DimString(const vector<int64_t>& dims) {
     if (dims.size() == 0) return "(0,)";
     std::stringstream ss;
@@ -187,7 +212,7 @@ class DRAGON_API Tensor {
     return name_;
   }
 
-  /*! \brief Return true if tensor name is set */
+  /*! \brief Return whether the tensor name is set */
   bool has_name() const {
     return !name_.empty();
   }
@@ -207,7 +232,7 @@ class DRAGON_API Tensor {
     return capacity_;
   }
 
-  /*! \brief Return the total number of bytes */
+  /*! \brief Return the total number of data bytes */
   size_t nbytes() const {
     return size_ * meta_.itemsize();
   }
@@ -231,50 +256,51 @@ class DRAGON_API Tensor {
     return (int)dims_.size();
   }
 
-  /*! \brief Return the dimension of specified axis */
+  /*! \brief Return the dimension of given axis */
   int64_t dim(int64_t i) const {
     return dims_[axis(i)];
   }
 
-  /*! \brief Return the stride of specified axis */
+  /*! \brief Return the stride of given axis */
   int64_t stride(int64_t i) const {
     return strides_[axis(i)];
   }
 
-  /*! \brief Return all the dimensions */
+  /*! \brief Return the tensor dimensions */
   const vec64_t& dims() const {
     return dims_;
   }
 
-  /*! \brief Return all the strides */
+  /*! \brief Return the tensor strides */
   const vec64_t& strides() const {
     return strides_;
   }
 
-  /*! \brief Return the number of elements along the [start, end) axes */
+  /*! \brief Return the total number of elements */
+  int64_t count() const {
+    return (int64_t)size_;
+  }
+
+  /*! \brief Return the number of elements counting along the given axes */
   int64_t count(int64_t start, int64_t end) const {
     int64_t nelements = 1;
-    for (int64_t i = start; i < end; i++)
+    for (int64_t i = start; i < end; i++) {
       nelements *= dim(i);
-    return nelements;
     }
-
-  /*! \brief Return the total number of elements */
-  int64_t count() const {
-    return (int64_t)size_;
+    return nelements;
   }
 
-  /*! \brief Return the number of elements from the start axis */
+  /*! \brief Return the number of elements counting from the given axis */
   int64_t count(int64_t start) const {
     return count(start, ndim());
   }
 
-  /*! \brief Whether this tensor is empty */
+  /*! \brief Return whether the total number of elements is zero */
   bool empty() const {
     return size_ == 0;
   }
 
-  /*! \brief Whether this tensor holds a valid memory */
+  /*! \brief Return whether the memory is set */
   bool has_memory() const {
     return internal_memory_ != nullptr || external_memory_ != nullptr;
   }
@@ -286,12 +312,12 @@ class DRAGON_API Tensor {
     return ptr;
   }
 
-  /*! \brief Return the state of memory */
+  /*! \brief Return the memory state */
   UnifiedMemory::State memory_state() const {
     return memory(true)->state();
   }
 
-  /*! \brief Try to get the raw const data pointer */
+  /*! \brief Try to return the raw const data pointer */
   template <class Context>
   const void* const_data_ptr() const {
     TypeId ctx_type = TypeMeta::Id<Context>();
@@ -307,7 +333,7 @@ class DRAGON_API Tensor {
     }
   }
 
-  /*! \brief Try to get the raw mutable data pointer */
+  /*! \brief Try to return the raw mutable data pointer */
   template <class Context>
   void mutable_data_ptr(void** data_ptr) {
     auto* mem = memory();
@@ -327,21 +353,23 @@ class DRAGON_API Tensor {
     }
   }
 
-  /*! \brief Try to allocate the raw data for memory */
+  /*!
+   * \brief Return the raw mutable data pointer.
+   *
+   * If memory is not set, create to manage it with the given meta.
+   */
   template <class Context>
   void* raw_mutable_data(const TypeMeta& meta) {
     void* data_ptr;
     mutable_data_ptr<Context>(&data_ptr);
-    // Return the data of memory directly
+    // Return the data pointer directly
     if (meta_ == meta && data_ptr) return data_ptr;
-    // Create a new memory with knowned size
+    // Create a new memory created with size and meta
     CHECK_GT(size_, 0) << "\nInvalid tensor size.";
     meta_ = meta;
     capacity_ = size_ * meta.itemsize();
     internal_memory_.reset(new UnifiedMemory(meta_, capacity_));
-    // Allocate space
     mutable_data_ptr<Context>(&data_ptr);
-    // Call the constructor if necessary
     if (meta_.ctor()) meta_.ctor()(data_ptr, size_);
     return data_ptr;
   }
@@ -360,7 +388,7 @@ class DRAGON_API Tensor {
     return const_data_ptr<Context>();
   }
 
-  /*! \brief Get the typed mutable data pointer */
+  /*! \brief Return the typed mutable data pointer */
   template <typename T, class Context>
   T* mutable_data() {
     void* data_ptr;
@@ -377,7 +405,7 @@ class DRAGON_API Tensor {
     return static_cast<T*>(raw_mutable_data<Context>(TypeMeta::Make<T>()));
   }
 
-  /*! \brief Get the typed const data pointer */
+  /*! \brief Return the typed const data pointer */
   template <typename T, class Context>
   const T* data() const {
     CHECK(meta_.Match<T>()) << "\nThe type of Tensor(" << name() << ") is "
@@ -391,13 +419,13 @@ class DRAGON_API Tensor {
     version_ = version;
   }
 
-  /*! \brief Set the meta of data type */
+  /*! \brief Set the type meta */
   Tensor* set_meta(const TypeMeta& meta) {
     meta_ = meta;
     return this;
   }
 
-  /*! \brief Set the internal memory */
+  /*! \brief Set to manage the memory */
   void set_memory(UnifiedMemory* memory) {
     if (memory != internal_memory_.get()) {
       internal_memory_.reset(memory);
@@ -429,6 +457,8 @@ class DRAGON_API Tensor {
 
   /*! \brief The external memory indicator */
   bool own_memory_ = true;
+
+  DISABLE_COPY_AND_ASSIGN(Tensor);
 };
 
 } // namespace dragon

dragon/core/typeid.h

@@ -31,15 +31,41 @@ struct DRAGON_API TypeRegister {
   }
 };
 
+/*!
+ * \brief Metaclass for all types.
+ *
+ * TypeMeta is commonly used for type identification:
+ *
+ * \code{.cpp}
+ * auto meta1 = dragon::TypeMeta::Make<float>();
+ * auto meta2 = dragon::TypeMeta::Make<float>();
+ * std::cout << (meta1 == meta2) << std::endl; // 1
+ * std::cout << (meta1.id() == meta2.id()) << std::endl; // 1
+ * std::cout << meta1.Match<float>() << std::endl; // 1
+ * std::cout << (meta1.id() == dragon::TypeMeta::Id<float>()) << std::endl; // 1
+ * \endcode
+ *
+ * Default constructor and destructor are available for non-fundamental types:
+ *
+ * \code{.cpp}
+ * auto meta = dragon::TypeMeta::Make<std::string>();
+ * auto* raw_string_data = malloc(1 * meta.itemsize());
+ * meta.ctor()(raw_string_data, 1);
+ * auto* string_data = reinterpret_cast<std::string*>(raw_string_data);
+ * std::cout << string_data[0].size();
+ * meta.dtor()(raw_string_data, 1);
+ * \endcode
+ */
 class TypeMeta {
  public:
   typedef void (*PlacementNew)(void*, size_t);
   typedef void (*TypedCopy)(const void*, void*, size_t);
   typedef void (*TypedDestructor)(void*, size_t);
 
-  TypeMeta()
-      : id_(0), itemsize_(0), ctor_(nullptr), copy_(nullptr), dtor_(nullptr) {}
+  /*! \brief Default constructor */
+  TypeMeta() : id_(0), itemsize_(0) {}
 
+  /*! \brief Constructor with the other type meta */
   TypeMeta(const TypeMeta& src)
       : id_(src.id_),
         itemsize_(src.itemsize_),
@@ -57,32 +83,38 @@ class TypeMeta {
     return *this;
   }
 
+  /*! \brief Return whether the two identifications are equal */
   bool operator==(const TypeMeta& other) const {
     return (id_ == other.id_);
   }
 
+  /*! \brief Return whether the two identifications are not equal */
   bool operator!=(const TypeMeta& other) const {
     return (id_ != other.id_);
   }
 
+  /*! \brief Return the identification of given type */
   template <typename T>
   static TypeId Id() {
     return TypeRegister<T>::id();
   }
 
+  /*! \brief Return the item size of given type */
   template <typename T>
   static size_t Itemsize() {
     return sizeof(T);
   }
 
+  /*! \brief Call the constructor for each element */
   template <typename T>
   static void Ctor(void* ptr, size_t n) {
     T* typed_ptr = static_cast<T*>(ptr);
-    for (size_t i = 0; i < n; i++) {
+    for (size_t i = 0; i < n; ++i) {
       new (typed_ptr + i) T;
     }
   }
 
+  /*! \brief Call the destructor for each element */
   template <typename T>
   static void Dtor(void* ptr, size_t n) {
     T* typed_ptr = static_cast<T*>(ptr);
@@ -91,55 +123,66 @@ class TypeMeta {
     }
   }
 
+  /*! \brief Call the copy constructor for each element */
   template <typename T>
   static void Copy(const void* src, void* dst, size_t n) {
     const T* typed_src = static_cast<const T*>(src);
     T* typed_dst = static_cast<T*>(dst);
-    for (size_t i = 0; i < n; ++i)
+    for (size_t i = 0; i < n; ++i) {
       typed_dst[i] = typed_src[i];
     }
+  }
 
-#define FundMeta std::enable_if<std::is_fundamental<T>::value, TypeMeta>::type
+#define FundamentalTypeMeta \
+  std::enable_if<std::is_fundamental<T>::value, TypeMeta>::type
 
-#define StructMeta                                                         \
+#define StructuralTypeMeta                                                 \
   std::enable_if<                                                          \
       !std::is_fundamental<T>::value && std::is_copy_assignable<T>::value, \
       TypeMeta>::type
 
+  /*! \brief Return a type meta of given type */
   template <typename T>
-  static typename FundMeta Make() {
+  static typename FundamentalTypeMeta Make() {
     return TypeMeta(Id<T>(), Itemsize<T>(), nullptr, nullptr, nullptr);
   }
 
+  /*! \brief Return a type meta of given type */
   template <typename T>
-  static typename StructMeta Make() {
+  static typename StructuralTypeMeta Make() {
     return TypeMeta(Id<T>(), Itemsize<T>(), Ctor<T>, Copy<T>, Dtor<T>);
   }
 
-#undef FundMeta
-#undef StructMeta
+#undef FundamentalTypeMeta
+#undef StructuralTypeMeta
 
+  /*! \brief Return whether the meta is matched with given type */
   template <typename T>
   bool Match() const {
     return (id_ == Id<T>());
   }
 
+  /*! \brief Return the type identification */
   const TypeId& id() const {
     return id_;
   }
 
+  /*! \brief Return the item size */
   const size_t& itemsize() const {
     return itemsize_;
   }
 
+  /*! \brief Return the type constructor */
   PlacementNew ctor() const {
     return ctor_;
   }
 
+  /*! \brief Return the type destructor */
   TypedDestructor dtor() const {
     return dtor_;
   }
 
+  /*! \brief Return the type copy constructor */
   TypedCopy copy() const {
     return copy_;
   }
@@ -156,9 +199,9 @@ class TypeMeta {
  private:
   TypeId id_;
   size_t itemsize_;
-  PlacementNew ctor_;
-  TypedCopy copy_;
-  TypedDestructor dtor_;
+  PlacementNew ctor_ = nullptr;
+  TypedCopy copy_ = nullptr;
+  TypedDestructor dtor_ = nullptr;
 };
 
 } // namespace dragon

dragon/core/workspace.cc

@@ -128,7 +128,7 @@ void Workspace::RunGraph(
     const int stream) {
   CHECK(graph_map_.count(name))
       << "\nGraph(" << name << ") is not in current workspace.";
-  graph_map_[name]->Run(include, exclude, stream);
+  graph_map_[name]->Run(stream, include, exclude);
 }
 
 void Workspace::RegisterAlias(const string& target, const string& alias) {

dragon/core/workspace.h

@@ -17,55 +17,58 @@
 
 namespace dragon {
 
-class Workspace {
+/*!
+ * \brief Sandbox to isolate the resources and computations.
+ */
+class DRAGON_API Workspace {
  public:
-  /*! \brief Constructor */
-  DRAGON_API explicit Workspace(const string& name);
+  /*! \brief Constructor with the name */
+  explicit Workspace(const string& name);
 
   /*! \brief Merge resources from other */
-  DRAGON_API void MergeFrom(Workspace*);
+  void MergeFrom(Workspace* other);
 
   /*! \brief Clear the cached resources */
-  DRAGON_API void Clear();
+  void Clear();
 
   /* \brief Return an unique name */
-  DRAGON_API string UniqueName(
+  string UniqueName(
       const string& name,
       const string& suffix,
       const string& scope = "",
       const bool zero_based = false);
 
   /* \brief Register an alias for the target */
-  DRAGON_API void RegisterAlias(const string& target, const string& alias);
+  void RegisterAlias(const string& target, const string& alias);
 
   /*! \brief Return whether tensor is existing */
-  DRAGON_API bool HasTensor(const string& name, bool external = true) const {
+  bool HasTensor(const string& name, bool external = true) const {
     return TryGetTensor(name, external) == nullptr ? false : true;
   }
 
   /*! \brief Create the tensor */
-  DRAGON_API Tensor* CreateTensor(const string&, FillerInfo* = nullptr);
+  Tensor* CreateTensor(const string&, FillerInfo* = nullptr);
 
   /*! \brief Try to return the tensor */
-  DRAGON_API Tensor* TryGetTensor(const string&, bool = true) const;
+  Tensor* TryGetTensor(const string&, bool = true) const;
 
   /*! \brief Return the tensor */
-  DRAGON_API Tensor* GetTensor(const string&, bool = true) const;
+  Tensor* GetTensor(const string&, bool = true) const;
 
   /*! \brief Reset the tensor */
-  DRAGON_API void ResetTensor(const string&);
+  void ResetTensor(const string&);
 
   /*! \brief Return the filler info */
-  DRAGON_API FillerInfo* GetFillerInfo(const string&);
+  FillerInfo* GetFillerInfo(const string&);
 
   /*! \brief Run the operator */
-  DRAGON_API void RunOperator(const OperatorDef&);
+  void RunOperator(const OperatorDef&);
 
   /*! \brief Create the graph */
-  DRAGON_API GraphBase* CreateGraph(const GraphDef&);
+  GraphBase* CreateGraph(const GraphDef&);
 
   /*! \brief Run the graph */
-  DRAGON_API void RunGraph(
+  void RunGraph(
       const string& graph_name,
       const string& include = "",
       const string& exclude = "",
@@ -77,12 +80,12 @@ class Workspace {
   }
 
   /*! \brief Return the name of cached tensors */
-  DRAGON_API vector<string> tensors() const;
+  vector<string> tensors() const;
 
   /*! \brief Return the name of cached graphs  */
-  DRAGON_API vector<string> graphs() const;
+  vector<string> graphs() const;
 
-  /*! \brief Provide a group of the shared byte data */
+  /*! \brief Return a group of the shared raw data */
   template <class Context>
   vector<void*> data(const vector<size_t>& segments) {
     int64_t nbytes = 0;
@@ -96,7 +99,7 @@ class Workspace {
     return ret;
   }
 
-  /*! \brief Provide a group of shared typed data */
+  /*! \brief Return a group of shared typed data */
   template <typename T, class Context>
   vector<T*> data(const vector<int64_t>& segments) {
     vector<size_t> segments_in_byte;
@@ -133,6 +136,8 @@ class Workspace {
 
   /*! \brief The cached graphs */
   Map<string, unique_ptr<GraphBase>> graph_map_;
+
+  DISABLE_COPY_AND_ASSIGN(Workspace);
 };
 
 } // namespace dragon

dragon/modules/python/common.h

@@ -69,7 +69,7 @@ class NumpyFetcher : public TensorFetcherBase {
           tensor.nbytes(),
           PyArray_DATA(reinterpret_cast<PyArrayObject*>(array)),
           tensor.raw_data<CUDAContext>(),
-          tensor.memory()->device_id());
+          tensor.memory()->device());
     } else {
       CPUContext::Memcpy<CPUContext, CPUContext>(
           tensor.nbytes(),

dragon/modules/python/cuda.h

@@ -130,7 +130,7 @@ void RegisterModule(py::module& m) {
 #ifdef USE_CUDA
     if (device_id < 0) device_id = CUDAContext::current_device();
     auto stream = CUDAContext::object()->stream(device_id, stream_id);
-    CUDAContext::SyncStream(stream);
+    CUDAContext::SynchronizeStream(stream);
 #endif
   });
 

dragon/modules/python/dlpack.h

@@ -50,7 +50,7 @@ class DLPackWrapper {
         } else {
           data = memory->mutable_cuda_data(nbytes);
         }
-        ctx.device_id = memory->device_id();
+        ctx.device_id = memory->device();
         ctx.device_type = DLDeviceType::kDLGPU;
         break;
       }

dragon/modules/python/module.cc

@@ -191,9 +191,12 @@ PYBIND11_MODULE(libdragon_python, m) {
             auto* graph = self->CreateGraph(graph_def);
             if (verbose) {
               bool could_be_serialized = true;
-              const auto& def = graph->opt_def();
-              for (auto& op : def.op())
-                if (op.type() == "GivenTensorFill") could_be_serialized = false;
+              const auto& def = graph->optimized_def();
+              for (auto& op : def.op()) {
+                if (op.type() == "GivenTensorFill") {
+                  could_be_serialized = false;
+                }
+              }
               if (could_be_serialized) {
                 auto msg = string("\n") + def.DebugString();
                 msg.pop_back();

dragon/modules/runtime/CMakeLists.txt

@@ -3,6 +3,9 @@ message(STATUS "Build module: ${CMAKE_CURRENT_LIST_DIR}")
 
 # ---[ Defines
 add_definitions(-DBUILD_RUNTIME)
+if (USE_MPI)
+  remove_definitions(-DUSE_MPI)
+endif()
 
 # ---[ Sources
 set(MODULE_INCLUDES "")

dragon/modules/runtime/device.cc

@@ -3,28 +3,33 @@
 
 namespace dragon {
 
-int type_from_string(std::string type) {
-  if (type == "CPU") {
+namespace {
+
+int type_from_string(const std::string& device_type) {
+  if (device_type == "CPU") {
     return 0;
-  } else if (type == "GPU") {
+  } else if (device_type == "GPU") {
     return 1;
-  } else if (type == "CUDA") {
+  } else if (device_type == "CUDA") {
     return 1;
   }
-  LOG(FATAL) << "Unknown device type: " << type << ", "
-             << "known device types: "
-             << "CPU, "
-             << "GPU, "
-             << "CUDA";
+  LOG(FATAL) << "Unsupported device type: " << device_type << "\n"
+             << "Following device types are supported: {"
+             << "  * CPU\n"
+             << "  * GPU\n"
+             << "  * CUDA\n"
+             << "}";
   return -1;
 }
 
+} // namespace
+
 Device::Device() : device_type_(0), device_id_(0) {}
 
-Device::Device(std::string device_type, int device_id)
+Device::Device(const std::string& device_type, int device_id)
     : device_type_(type_from_string(device_type)), device_id_(device_id) {}
 
-Device::Device(std::string device_type)
+Device::Device(const std::string& device_type)
     : device_type_(type_from_string(device_type)), device_id_(0) {}
 
 } // namespace dragon

dragon/modules/runtime/dragon_runtime.h

@@ -40,8 +40,8 @@ typedef class Workspace* Workspace_t;
 class DRAGON_API Device {
  public:
   Device();
-  explicit Device(std::string device_type);
-  Device(std::string device_type, int device_id);
+  explicit Device(const std::string& device_type);
+  Device(const std::string& device_type, int device_id);
 
   const int& device_type() const {
     return device_type_;
@@ -65,7 +65,7 @@ DRAGON_API Workspace_t ResetWorkspace(Workspace_t ws);
 
 DRAGON_API Workspace_t ResetWorkspace(const std::string& name);
 
-DRAGON_API void MoveWorkspace(Workspace_t dst, Workspace_t src);
+DRAGON_API void MoveWorkspace(Workspace_t dest, Workspace_t src);
 
 DRAGON_API void DestroyWorkspace(Workspace_t ws);
 
@@ -76,15 +76,13 @@ DRAGON_API void DestroyWorkspace(const std::string& name);
  */
 
 DRAGON_API std::string
-CreateGraph(const GraphDef_t graph_def, const Device& device, Workspace_t ws);
+CreateGraph(const GraphDef_t def, const Device& device, Workspace_t ws);
 
-DRAGON_API std::string CreateGraph(
-    const std::string& graph_file,
-    const Device& device,
-    Workspace_t ws);
+DRAGON_API std::string
+CreateGraph(const std::string& file, const Device& device, Workspace_t ws);
 
 DRAGON_API void
-RunGraph(const std::string& graph_name, Workspace_t ws, int stream_id = 0);
+RunGraph(const std::string& name, Workspace_t ws, int stream = 0);
 
 /*!
  * Tensor API
@@ -111,9 +109,9 @@ DRAGON_API T* FetchTensor(
  * Proto API
  */
 
-DRAGON_API void CreateGraphDef(GraphDef_t* graph_def);
+DRAGON_API void CreateGraphDef(GraphDef_t* def);
 
-DRAGON_API void DestroyGraphDef(GraphDef_t graph_def);
+DRAGON_API void DestroyGraphDef(GraphDef_t def);
 
 /*!
  * Model API
@@ -121,8 +119,8 @@ DRAGON_API void DestroyGraphDef(GraphDef_t graph_def);
 
 DRAGON_API void LoadONNXModel(
     const std::string& model_file,
-    GraphDef_t init_graph,
-    GraphDef_t pred_graph,
+    GraphDef_t init_def,
+    GraphDef_t pred_def,
     std::vector<std::string>& inputs,
     std::vector<std::string>& outputs);
 

dragon/modules/runtime/module.cc

-#include "dragon/core/common.h"
+#include "dragon/core/workspace.h"
 #include "dragon/modules/runtime/dragon_runtime.h"
 #include "dragon/onnx/onnx_backend.h"
 #include "dragon/utils/proto_utils.h"
@@ -9,7 +9,7 @@ std::mutex g_mutex;
 Map<string, unique_ptr<Workspace>> g_workspaces;
 Map<string, vector<string>> sub_workspaces;
 
-Workspace* CreateWorkspace(const string& name) {
+Workspace_t CreateWorkspace(const string& name) {
   std::unique_lock<std::mutex> lock(g_mutex);
   LOG(INFO) << "Create the Workspace(" << name << ").";
   if (g_workspaces.count(name)) return g_workspaces[name].get();
@@ -19,7 +19,7 @@ Workspace* CreateWorkspace(const string& name) {
   return g_workspaces[name].get();
 }
 
-Workspace* ResetWorkspace(const string& name) {
+Workspace_t ResetWorkspace(const string& name) {
   std::unique_lock<std::mutex> lock(g_mutex);
   CHECK(g_workspaces.count(name))
       << "\nWorkspace(" << name << ") does not exist."
@@ -34,19 +34,19 @@ Workspace* ResetWorkspace(const string& name) {
   return g_workspaces[name].get();
 }
 
-Workspace* ResetWorkspace(Workspace_t ws) {
+Workspace_t ResetWorkspace(Workspace_t ws) {
   CHECK(ws) << "\nGiven workspace is invalid.";
   return ResetWorkspace(ws->name());
 }
 
-void MoveWorkspace(Workspace_t dst, Workspace_t src) {
+void MoveWorkspace(Workspace_t dest, Workspace_t src) {
   std::unique_lock<std::mutex> lock(g_mutex);
   CHECK(src) << "\nGiven source workspace is invalid.";
-  CHECK(dst) << "\nGiven destination workspace is invalid.";
-  dst->MergeFrom(src);
-  sub_workspaces[dst->name()].push_back(src->name());
+  CHECK(dest) << "\nGiven destination workspace is invalid.";
+  dest->MergeFrom(src);
+  sub_workspaces[dest->name()].push_back(src->name());
   LOG(INFO) << "Move the Workspace(" << src->name() << ") "
-            << "into the Workspace(" << dst->name() << ").";
+            << "into the Workspace(" << dest->name() << ").";
 }
 
 void DestroyWorkspace(const string& name) {
@@ -63,27 +63,25 @@ void DestroyWorkspace(Workspace_t ws) {
   return DestroyWorkspace(ws->name());
 }
 
-string
-CreateGraph(const GraphDef_t graph_def, const Device& device, Workspace_t ws) {
-  auto graph_def_copy(*graph_def);
-  // Overwritten device options
-  DeviceOption* device_option = graph_def_copy.mutable_device_option();
+string CreateGraph(const GraphDef_t def, const Device& device, Workspace_t ws) {
+  auto def_v2(*def);
+  auto* device_option = def_v2.mutable_device_option();
   device_option->set_device_type((DeviceTypeProto)device.device_type());
   device_option->set_device_id(device.device_id());
-  auto* graph = ws->CreateGraph(graph_def_copy);
+  auto* graph = ws->CreateGraph(def_v2);
   if (!graph) LOG(FATAL) << "Can not create the graph.";
   return graph->name();
 }
 
 std::string
-CreateGraph(const string& graph_file, const Device& device, Workspace_t ws) {
+CreateGraph(const string& file, const Device& device, Workspace_t ws) {
   GraphDef graph_def;
-  ParseProtoFromText(graph_file.c_str(), &graph_def);
+  ParseProtoFromText(file.c_str(), &graph_def);
   return CreateGraph(&graph_def, device, ws);
 }
 
-void RunGraph(const string& graph_name, Workspace_t ws, const int stream_id) {
-  ws->RunGraph(graph_name, "", "", stream_id);
+void RunGraph(const string& name, Workspace_t ws, int stream) {
+  ws->RunGraph(name, "", "", stream);
 }
 
 void CreateTensor(const string& name, Workspace_t ws) {
@@ -148,34 +146,38 @@ void FeedTensor(
         tensor->raw_mutable_data<CPUContext>(),
         static_cast<const void*>(data));
   } else {
-    LOG(FATAL) << "Unknown device type.";
+    LOG(FATAL) << "Unsupported device type.";
   }
 }
 
-DRAGON_API void CreateGraphDef(GraphDef_t* graph_def) {
-  *graph_def = new GraphDef();
+void CreateGraphDef(GraphDef_t* def) {
+  *def = new GraphDef();
 }
 
-DRAGON_API void DestroyGraphDef(GraphDef_t graph_def) {
-  if (graph_def) delete graph_def;
+void DestroyGraphDef(GraphDef_t def) {
+  if (def) {
+    delete def;
+  }
 }
 
 void LoadONNXModel(
     const string& model_file,
-    GraphDef_t init_graph,
-    GraphDef_t pred_graph,
+    GraphDef_t init_def,
+    GraphDef_t pred_def,
     vector<string>& inputs,
     vector<string>& outputs) {
   LOG(INFO) << "Load Model: " << model_file << "......";
   LOG(INFO) << "Format: ONNX";
   onnx::ONNXBackend onnx_backend;
-  onnx_backend.Prepare(model_file, init_graph, pred_graph);
+  onnx_backend.Prepare(model_file, init_def, pred_def);
   inputs.clear();
   outputs.clear();
-  for (const auto& e : pred_graph->input())
-    inputs.emplace_back(e);
-  for (const auto& e : pred_graph->output())
-    outputs.emplace_back(e);
+  for (const auto& input : pred_def->input()) {
+    inputs.push_back(input);
+  }
+  for (const auto& output : pred_def->output()) {
+    outputs.push_back(output);
+  }
 }
 
 #define INSTANTIATE_API(T)                \

dragon/operators/array/cast_op.cc

@@ -8,54 +8,56 @@ namespace dragon {
 #define ELIGIBLE_TENSOR_TYPES \
   { "bool", "int8", "uint8", "int32", "int64", "float16", "float32", "float64" }
 
-#define DEFINE_TYPE_A_TO_B(Ta, type_str, Tb)                     \
-  if (dtype() == type_str) {                                     \
+#define DISPATCH_TYPE_TO(InputType, OutputType)                          \
+  if (dtype() == types::to_string<OutputType>()) {                       \
     if (InputSize() != 0) {                                              \
       Output(0)->ReshapeLike(Input(0));                                  \
-      auto* x = Input(0).template data<Ta, Context>();           \
-      auto* y = Output(0)->template mutable_data<Tb, Context>(); \
+      auto* x = Input(0).template data<InputType, Context>();            \
+      auto* y = Output(0)->template mutable_data<OutputType, Context>(); \
       kernel::Cast(Input(0).count(), x, y, ctx());                       \
     } else {                                                             \
       auto n = Output(0)->count();                                       \
-      auto* x = Output(0)->template data<Ta, Context>();         \
-      auto* scratch = ws()->template data<Tb, Context>({n})[0];  \
+      auto* x = Output(0)->template data<InputType, Context>();          \
+      auto* scratch = ws()->template data<OutputType, Context>({n})[0];  \
       kernel::Cast(n, x, scratch, ctx());                                \
       ctx()->FinishDeviceComputation();                                  \
-      auto* y = Output(0)->template mutable_data<Tb, Context>(); \
+      auto* y = Output(0)->template mutable_data<OutputType, Context>(); \
       math::Copy(n, scratch, y, ctx());                                  \
     }                                                                    \
     return;                                                              \
   }
 
-#define DEFINE_TYPE_A_TO_ALL(Ta)              \
-  DEFINE_TYPE_A_TO_B(Ta, "bool", bool);       \
-  DEFINE_TYPE_A_TO_B(Ta, "int8", int8_t);     \
-  DEFINE_TYPE_A_TO_B(Ta, "uint8", uint8_t);   \
-  DEFINE_TYPE_A_TO_B(Ta, "int32", int);       \
-  DEFINE_TYPE_A_TO_B(Ta, "int64", int64_t);   \
-  DEFINE_TYPE_A_TO_B(Ta, "float16", float16); \
-  DEFINE_TYPE_A_TO_B(Ta, "float32", float);   \
-  DEFINE_TYPE_A_TO_B(Ta, "float64", double)
+#define DISPATCH_TYPE_TO_ALL(InputType) \
+  DISPATCH_TYPE_TO(InputType, bool);    \
+  DISPATCH_TYPE_TO(InputType, int8_t);  \
+  DISPATCH_TYPE_TO(InputType, uint8_t); \
+  DISPATCH_TYPE_TO(InputType, int);     \
+  DISPATCH_TYPE_TO(InputType, int64_t); \
+  DISPATCH_TYPE_TO(InputType, float16); \
+  DISPATCH_TYPE_TO(InputType, float);   \
+  DISPATCH_TYPE_TO(InputType, double);  \
+  LOG(FATAL) << MessageForUnsupported(dtype(), ELIGIBLE_TENSOR_TYPES);
 
 #define DISPATCH_WITH_TENSOR(X)                             \
   if (XIsType(X, bool)) {                                   \
-    DEFINE_TYPE_A_TO_ALL(bool);                         \
+    DISPATCH_TYPE_TO_ALL(bool);                             \
   } else if (XIsType(X, int8_t)) {                          \
-    DEFINE_TYPE_A_TO_ALL(int8_t);                       \
+    DISPATCH_TYPE_TO_ALL(int8_t);                           \
   } else if (XIsType(X, uint8_t)) {                         \
-    DEFINE_TYPE_A_TO_ALL(uint8_t);                      \
+    DISPATCH_TYPE_TO_ALL(uint8_t);                          \
   } else if (XIsType(X, int)) {                             \
-    DEFINE_TYPE_A_TO_ALL(int);                          \
+    DISPATCH_TYPE_TO_ALL(int);                              \
   } else if (XIsType(X, int64_t)) {                         \
-    DEFINE_TYPE_A_TO_ALL(int64_t);                      \
+    DISPATCH_TYPE_TO_ALL(int64_t);                          \
   } else if (XIsType(X, float16)) {                         \
-    DEFINE_TYPE_A_TO_ALL(float16);                      \
+    DISPATCH_TYPE_TO_ALL(float16);                          \
   } else if (XIsType(X, float)) {                           \
-    DEFINE_TYPE_A_TO_ALL(float);                        \
+    DISPATCH_TYPE_TO_ALL(float);                            \
   } else if (XIsType(X, double)) {                          \
-    DEFINE_TYPE_A_TO_ALL(double);                       \
+    DISPATCH_TYPE_TO_ALL(double);                           \
   } else {                                                  \
-    LOG(FATAL) << TypeString(X, ELIGIBLE_TENSOR_TYPES); \
+    LOG(FATAL) << MessageForUnsupported(                    \
+        types::to_string(X.meta()), ELIGIBLE_TENSOR_TYPES); \
   }
 
 template <class Context>
@@ -101,8 +103,8 @@ OPERATOR_SCHEMA(CastGradient)
 REGISTER_GRADIENT(Cast, SimpleGradientMaker);
 
 #undef ELIGIBLE_TENSOR_TYPES
-#undef DEFINE_TYPE_A_TO_B
-#undef DEFINE_TYPE_A_TO_ALL
+#undef DISPATCH_TYPE_TO
+#undef DISPATCH_TYPE_TO_ALL
 #undef DISPATCH_WITH_TENSOR
 
 } // namespace dragon

dragon/operators/array/channel_normalize_op.cc

@@ -52,7 +52,8 @@ void ChannelNormalizeOp<Context>::DoRunWithType() {
   } else if (dtype() == "float64") {
     DoRunWithTypeAndCast<T, double>();
   } else {
-    LOG(FATAL) << TypeString(dtype(), {"float16", "float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        dtype(), {"float16", "float32", "float64"});
   }
 }
 

dragon/operators/array/multinomial_op.cc

@@ -60,7 +60,7 @@ void MultinomialOp<Context>::DoRunWithType() {
 
 template <class Context>
 void MultinomialOp<Context>::RunOnDevice() {
-  ctx()->set_stream_id(0); // Enforce the default stream
+  ctx()->set_stream(0); // Enforce the default stream
   DispatchHelper<TensorTypes<float, double>>::Call(this, Input(0));
 }
 

dragon/operators/distributed/collective_op.cc

@@ -124,23 +124,24 @@ template <class Context>
 void CollectiveOp<Context>::RunOnDevice() {
   if (communication_ == "ALLREDUCE") {
     for (int i = 0; i < InputSize(); i++) {
-      if (XIsType(Input(i), int8_t)) {
+      auto& X = Input(i);
+      if (XIsType(X, int8_t)) {
         AllReduceDispatcher<int8_t>(&Input(i));
-      } else if (XIsType(Input(i), uint8_t)) {
+      } else if (XIsType(X, uint8_t)) {
         AllReduceDispatcher<uint8_t>(&Input(i));
-      } else if (XIsType(Input(i), int)) {
+      } else if (XIsType(X, int)) {
         AllReduceDispatcher<int>(&Input(i));
-      } else if (XIsType(Input(i), int64_t)) {
+      } else if (XIsType(X, int64_t)) {
         AllReduceDispatcher<int64_t>(&Input(i));
-      } else if (XIsType(Input(i), float16)) {
+      } else if (XIsType(X, float16)) {
         AllReduceDispatcher<float16>(&Input(i));
-      } else if (XIsType(Input(i), float)) {
+      } else if (XIsType(X, float)) {
         AllReduceDispatcher<float>(&Input(i));
-      } else if (XIsType(Input(i), double)) {
+      } else if (XIsType(X, double)) {
         AllReduceDispatcher<double>(&Input(i));
       } else {
-        LOG(FATAL) << TypeString(
-            Input(i),
+        LOG(FATAL) << MessageForUnsupported(
+            types::to_string(X.meta()),
             {"int8",
              "uint8",
              "int32",
@@ -152,25 +153,26 @@ void CollectiveOp<Context>::RunOnDevice() {
     }
   } else if (communication_ == "BROADCAST") {
     for (int i = 0; i < InputSize(); i++) {
-      if (XIsType(Input(i), bool)) {
+      auto& X = Input(i);
+      if (XIsType(X, bool)) {
         BroadcastDispatcher<bool>(&Input(i));
-      } else if (XIsType(Input(i), int8_t)) {
+      } else if (XIsType(X, int8_t)) {
         BroadcastDispatcher<int8_t>(&Input(i));
-      } else if (XIsType(Input(i), uint8_t)) {
+      } else if (XIsType(X, uint8_t)) {
         BroadcastDispatcher<uint8_t>(&Input(i));
-      } else if (XIsType(Input(i), int)) {
+      } else if (XIsType(X, int)) {
         BroadcastDispatcher<int>(&Input(i));
-      } else if (XIsType(Input(i), int64_t)) {
+      } else if (XIsType(X, int64_t)) {
         BroadcastDispatcher<int64_t>(&Input(i));
-      } else if (XIsType(Input(i), float16)) {
+      } else if (XIsType(X, float16)) {
         BroadcastDispatcher<float16>(&Input(i));
-      } else if (XIsType(Input(i), float)) {
+      } else if (XIsType(X, float)) {
         BroadcastDispatcher<float>(&Input(i));
-      } else if (XIsType(Input(i), double)) {
+      } else if (XIsType(X, double)) {
         BroadcastDispatcher<double>(&Input(i));
       } else {
-        LOG(FATAL) << TypeString(
-            Input(i),
+        LOG(FATAL) << MessageForUnsupported(
+            types::to_string(X.meta()),
             {"bool",
              "int8",
              "uint8",

dragon/operators/distributed/collective_op_base.h

@@ -149,7 +149,7 @@ class CollectiveOpBase : public Operator<Context> {
 
   ncclComm_t nccl_comm() {
     auto ret = CUDAContext::object()->nccl_comm(
-        this->ctx()->template device_id(),
+        this->ctx()->template device(),
         group_str_,
         nullptr,
         comm_size_,
@@ -162,7 +162,7 @@ class CollectiveOpBase : public Operator<Context> {
       }
       Broadcast((uint8_t*)&comm_uuid, sizeof(comm_uuid));
       ret = CUDAContext::object()->nccl_comm(
-          this->ctx()->template device_id(),
+          this->ctx()->template device(),
           group_str_,
           &comm_uuid,
           comm_size_,

dragon/operators/loss/ctc_loss_op_cudnn.cc

@@ -85,7 +85,8 @@ void CuDNNCTCLossOp<Context>::RunOnDevice() {
     CUDNN_CHECK(cudnnSetCTCLossDescriptor(ctc_desc_, CUDNN_DATA_FLOAT));
     DoRunWithType<float>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 

dragon/operators/loss/nll_loss_op.cc

@@ -72,7 +72,8 @@ void NLLLossOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float32", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -80,10 +81,12 @@ void NLLLossOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float64", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 
@@ -139,7 +142,8 @@ void NLLLossGradientOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float32", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -147,10 +151,12 @@ void NLLLossGradientOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float64", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 

dragon/operators/loss/sigmoid_focal_loss_op.cc

@@ -72,7 +72,8 @@ void SigmoidFocalLossOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float32", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -80,10 +81,12 @@ void SigmoidFocalLossOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float64", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 
@@ -139,7 +142,8 @@ void SigmoidFocalLossGradientOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float32", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -147,10 +151,12 @@ void SigmoidFocalLossGradientOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float64", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 

dragon/operators/loss/sparse_softmax_ce_loss_op.cc

@@ -82,7 +82,8 @@ void SparseSoftmaxCrossEntropyOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float32", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -90,10 +91,12 @@ void SparseSoftmaxCrossEntropyOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float64", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 
@@ -152,7 +155,8 @@ void SparseSoftmaxCrossEntropyGradientOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float32", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -160,10 +164,12 @@ void SparseSoftmaxCrossEntropyGradientOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"float64", "int64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 

dragon/operators/math/axpby_op.cc

@@ -45,8 +45,8 @@ void AxpbyOp<Context>::RunOnDevice() {
     } else if (XIsType(X, double)) {
       DoRunWithType<double>(&X, Y);
     } else
-      LOG(FATAL) << TypeString(
-          X,
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(X.meta()),
           {"int8", "uint8", "int32", "int64", "float16", "float32", "float64"});
   }
 }

dragon/operators/math/moments_op.cc

@@ -75,8 +75,8 @@ void MomentsOp<Context>::RunOnDevice() {
   } else if (XIsType(X, double)) {
     DoRunWithType<double, double>();
   } else {
-    LOG(FATAL) << TypeString(
-        X,
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(X.meta()),
         {"int8", "uint8", "int32", "int64", "float16", "float32", "float64"});
   }
 }

dragon/operators/metric/accuracy_op.cc

@@ -55,7 +55,8 @@ void AccuracyOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<float, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"int64", "float32"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float32", "int64"});
     }
   } else if (XIsType(Input(0), double)) {
     if (XIsType(Input(1), double)) {
@@ -63,10 +64,12 @@ void AccuracyOp<Context>::RunOnDevice() {
     } else if (XIsType(Input(1), int64_t)) {
       DoRunWithType<double, int64_t>();
     } else {
-      LOG(FATAL) << TypeString(Input(1), {"int64", "float64"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(1).meta()), {"float64", "int64"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32", "float64"});
   }
 }
 

dragon/operators/normalization/batch_norm_op.cc

@@ -114,7 +114,8 @@ void BatchNormOp<Context>::RunOnDevice() {
       InferenceImpl<float, float>();
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 
@@ -190,7 +191,8 @@ void BatchNormGradientOp<Context>::RunOnDevice() {
       InferenceImpl<float, float>();
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 

dragon/operators/normalization/batch_norm_op_cudnn.cc

@@ -90,7 +90,8 @@ void CuDNNBatchNormOp<Context>::RunOnDevice() {
   } else if (XIsType(Input(0), float16)) {
     DoRunWithType<float16>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float16"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float16", "float32"});
   }
 }
 
@@ -156,10 +157,12 @@ void CuDNNBatchNormGradientOp<Context>::RunOnDevice() {
       TrainingImpl<float16>();
     } else {
       // We will support it some day -:)
-      LOG(FATAL) << TypeString(Input(0), {"float32"});
+      LOG(FATAL) << MessageForUnsupported(
+          types::to_string(Input(0).meta()), {"float32"});
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float16", "float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float16", "float32"});
   }
 }
 

dragon/operators/normalization/batch_norm_op_sync.cc

@@ -111,7 +111,8 @@ void SyncBatchNormOp<Context>::RunOnDevice() {
       this->template InferenceImpl<float, float>();
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 
@@ -195,7 +196,8 @@ void SyncBatchNormGradientOp<Context>::RunOnDevice() {
       this->template InferenceImpl<float, float>();
     }
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 

dragon/operators/normalization/group_norm_op.cc

@@ -60,7 +60,8 @@ void GroupNormOp<Context>::RunOnDevice() {
   } else if (XIsType(Input(0), float16)) {
     DoRunWithType<float16, float>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32", "float16"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float16", "float32"});
   }
 }
 
@@ -101,7 +102,8 @@ void GroupNormGradientOp<Context>::RunOnDevice() {
   } else if (XIsType(Input(0), float16)) {
     DoRunWithType<float16, float>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float16", "float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float16", "float32"});
   }
 }
 

dragon/operators/recurrent/lstm_cell_op.cc

@@ -23,7 +23,8 @@ void LSTMCellOp<Context>::RunOnDevice() {
   if (XIsType(Input(0), float)) {
     DoRunWithType<float>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 
@@ -60,7 +61,8 @@ void LSTMCellGradientOp<Context>::RunOnDevice() {
   if (XIsType(Input(0), float)) {
     DoRunWithType<float>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float32"});
   }
 }
 

dragon/operators/training/update_op_base.cc

@@ -100,7 +100,8 @@ void UpdateOpBase<Context>::RunOnDevice() {
     ComputeUpdate(dX_cast);
     ApplyUpdate<float>(dX_cast, X);
   } else {
-    LOG(FATAL) << TypeString(dX, {"float16", "float32"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(dX.meta()), {"float16", "float32"});
   }
 }
 

dragon/operators/vision/resize_op.cc

@@ -199,7 +199,8 @@ void ResizeGradientOp<Context>::RunOnDevice() {
   } else if (XIsType(Input(0), double)) {
     DoRunWithTypeAndCast<double>();
   } else {
-    LOG(FATAL) << TypeString(Input(0), {"float16", "float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(0).meta()), {"float16", "float32", "float64"});
   };
 }
 

dragon/operators/vision/roi_align_op.cc

@@ -95,7 +95,8 @@ void RoiAlignGradientOp<Context>::RunOnDevice() {
   } else if (XIsType(Input(1), double)) {
     DoRunWithTypeAndCast<double>();
   } else {
-    LOG(FATAL) << TypeString(Input(1), {"float16", "float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(1).meta()), {"float16", "float32", "float64"});
   };
 }
 

dragon/operators/vision/roi_pool_op.cc

@@ -98,7 +98,8 @@ void RoiPoolGradientOp<Context>::RunOnDevice() {
   } else if (XIsType(Input(1), double)) {
     DoRunWithTypeAndCast<double>();
   } else {
-    LOG(FATAL) << TypeString(Input(1), {"float16", "float32", "float64"});
+    LOG(FATAL) << MessageForUnsupported(
+        types::to_string(Input(1).meta()), {"float16", "float32", "float64"});
   };
 }
 

dragon/python/core/framework/context.py

@@ -52,14 +52,14 @@ def device(device_type, device_index=0):
 
 
 def eager_scope(data='${DATA}', graph='${GRAPH}'):
-    """Context-manager to nest the domain for eager resources.
+    """Context-manager to nest the namespace for eager resources.
 
     Parameters
     ----------
     data : str, optional, default='${DATA}'
-        The domain for resources traced by python.
+        The namespace for resources traced by python.
     graph : str, optional, default='${GRAPH}'
-        The domain for resources traced by graph.
+        The namespace for resources traced by graph.
 
     """
     domain_tuple = (graph, data)

dragon/python/core/framework/workspace.py

@@ -105,7 +105,7 @@ class Workspace(backend.Workspace):
         return self._collectors
 
     def as_default(self):
-        """Switch ``self`` as the default workspace.
+        """Switch this workspace as the default.
 
         Call this method with the **with** keyword.
 
@@ -114,7 +114,7 @@ class Workspace(backend.Workspace):
         Returns
         -------
         dragon.Workspace
-            The ``self``.
+            This workspace.
 
         """
         return _GLOBAL_DEFAULT_WORKSPACE_STACK.get_controller(self)
@@ -273,7 +273,7 @@ class Workspace(backend.Workspace):
         Returns
         -------
         dragon.Workspace
-            The ``self``.
+            This workspace.
 
         """
         self.MergeFrom(other)
@@ -302,7 +302,7 @@ class Workspace(backend.Workspace):
             The tensor to reset.
 
         """
-        return self.ResetTensor(_stringify_object(tensor))
+        self.ResetTensor(_stringify_object(tensor))
 
     def run_backward(
         self,
@@ -487,8 +487,7 @@ _GLOBAL_DEFAULT_WORKSPACE_STACK = _DefaultWorkspaceStack()
 # Predefined graph executing stages.
 _PREDEFINED_GRAPH_EXECUTING_STAGES = {
     'default': {'include': '', 'exclude': ''},
-    'forward': {'include': '', 'exclude': 'Gradient'},
-    'backward': {'include': 'Gradient', 'exclude': 'Generate'},
-    'backward_v2': {'include': 'Gradient', 'exclude': ''},
-    'external_grads': {'include': '', 'exclude': 'Generate'},
+    'forward': {'include': '', 'exclude': '.*Gradient.*'},
+    'backward': {'include': '.*Gradient.*', 'exclude': 'GradientGenerate'},
+    'backward_v2': {'include': '.*Gradient.*', 'exclude': ''},
 }

dragon/setup.py

@@ -153,7 +153,7 @@ setuptools.setup(
     package_data={'dragon': find_package_data()},
     package_dir={'dragon': 'dragon'},
     cmdclass={'bdist_wheel': bdist_wheel, 'install': install},
-    python_requires='>=2.7, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*',
+    python_requires='>=3.5',
     install_requires=['numpy', 'protobuf', 'kpl-dataset'],
     classifiers=[
         'Development Status :: 5 - Production/Stable',
@@ -162,12 +162,12 @@ setuptools.setup(
         'Intended Audience :: Science/Research',
         'License :: OSI Approved :: BSD License',
         'Programming Language :: C++',
-        'Programming Language :: Python :: 2',
-        'Programming Language :: Python :: 2.7',
         'Programming Language :: Python :: 3',
+        'Programming Language :: Python :: 3 :: Only',
         'Programming Language :: Python :: 3.5',
         'Programming Language :: Python :: 3.6',
         'Programming Language :: Python :: 3.7',
+        'Programming Language :: Python :: 3.8',
         'Topic :: Scientific/Engineering',
         'Topic :: Scientific/Engineering :: Mathematics',
         'Topic :: Scientific/Engineering :: Artificial Intelligence',

dragon/utils/macros.h

@@ -24,10 +24,19 @@
 #define TLS_OBJECT __declspec(thread)
 #endif
 
+// Disable the copy and assignment operator for a class
+#define DISABLE_COPY_AND_ASSIGN(classname) \
+  classname(const classname&) = delete;    \
+  classname& operator=(const classname&) = delete
+
+// Concatenate two strings
 #define CONCATENATE_IMPL(s1, s2) s1##s2
 #define CONCATENATE(s1, s2) CONCATENATE_IMPL(s1, s2)
+
+// Return a anonymous variable name using line number
 #define ANONYMOUS_VARIABLE(str) CONCATENATE(str, __LINE__)
-#define NOT_IMPLEMENTED \
-  LOG(FATAL) << "This module has not been implemented yet."
+
+// Throw a fatal logging for not implemented function
+#define NOT_IMPLEMENTED LOG(FATAL) << "This function is not implemented."
 
 #endif // DRAGON_UTILS_MARCROS_H_

dragon/version.txt

-0.3.0
+0.3.0a0

scripts/build_windows_vc12.bat

-:: #############################################################################
-:: Example command to build on Windows for Visual Studio 2013 (VC12).
-:: #############################################################################
+:: ##############################################################
+:: Command file to build on Windows for Visual Studio 2013 (VC12)
+:: ##############################################################
 
 @echo off
 setlocal
 
-SET ORIGINAL_DIR=%cd%
-SET REPO_ROOT=%~dp0%..
-SET DRAGON_ROOT=%REPO_ROOT%\dragon
-SET THIRD_PARTY_DIR=%REPO_ROOT%\third_party
-SET CMAKE_GENERATOR="Visual Studio 12 2013 Win64"
+:: Build variables
+set ORIGINAL_DIR=%cd%
+set REPO_ROOT=%~dp0%..
+set DRAGON_ROOT=%REPO_ROOT%\dragon
+set THIRD_PARTY_DIR=%REPO_ROOT%\third_party
+set CMAKE_GENERATOR="Visual Studio 12 2013 Win64"
 
 :: Build options
-SET BUILD_PYTHON=ON
-SET BUILD_RUNTIME=OFF
+set BUILD_PYTHON=ON
+set BUILD_RUNTIME=OFF
+
+:: Optional libraries
+set USE_CUDA=ON
+set USE_CUDNN=ON
+set USE_OPENMP=ON
+set USE_AVX=ON
+set USE_AVX2=ON
+set USE_FMA=ON
 
 :: Protobuf SDK options
-SET PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
+set PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
 
 :: Protobuf Compiler options
-:: Set the protobuf compiler(i.e., protoc) if necessary
-:: If not, a compiler in the sdk or environment will be used
-SET PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
+:: Set the protobuf compiler(i.e., protoc) if necessary.
+:: If not, a compiler in the sdk or environment will be used.
+set PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
 
 :: Python options
-:: Set your python "interpreter" if necessary
-:: If not, a default interpreter will be used
-:: SET PYTHON_EXECUTABLE=X:/Anaconda3/python
+:: Set your python "interpreter" if necessary.
+:: If not, a default interpreter will be used.
+:: set PYTHON_EXECUTABLE=X:/Anaconda3/python
+
 if %BUILD_PYTHON% == ON (
   if NOT DEFINED PYTHON_EXECUTABLE (
     for /F %%i in ('python -c "import sys;print(sys.executable)"') do (set PYTHON_EXECUTABLE=%%i)
@@ -47,6 +57,12 @@ cmake .. ^
   -G%CMAKE_GENERATOR% ^
   -DBUILD_PYTHON=%BUILD_PYTHON% ^
   -DBUILD_RUNTIME=%BUILD_RUNTIME% ^
+  -USE_CUDA==%USE_CUDA% ^
+  -USE_CUDNN==%USE_CUDNN% ^
+  -USE_OPENMP==%USE_OPENMP% ^
+  -USE_AVX==%USE_AVX% ^
+  -USE_AVX2==%USE_AVX2% ^
+  -USE_FMA==%USE_FMA% ^
   -DTHIRD_PARTY_DIR=%THIRD_PARTY_DIR% ^
   -DPROTOBUF_SDK_ROOT_DIR=%PROTOBUF_SDK_ROOT_DIR% ^
   -DPROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_PROTOC_EXECUTABLE% ^

scripts/build_windows_vc14.bat

-:: #############################################################################
-:: Example command to build on Windows for Visual Studio 2015 (VC14).
-:: #############################################################################
+:: ##############################################################
+:: Command file to build on Windows for Visual Studio 2015 (VC14)
+:: ##############################################################
 
 @echo off
 setlocal
 
-SET ORIGINAL_DIR=%cd%
-SET REPO_ROOT=%~dp0%..
-SET DRAGON_ROOT=%REPO_ROOT%\dragon
-SET THIRD_PARTY_DIR=%REPO_ROOT%\third_party
-SET CMAKE_GENERATOR="Visual Studio 14 2015 Win64"
+:: Build variables
+set ORIGINAL_DIR=%cd%
+set REPO_ROOT=%~dp0%..
+set DRAGON_ROOT=%REPO_ROOT%\dragon
+set THIRD_PARTY_DIR=%REPO_ROOT%\third_party
+set CMAKE_GENERATOR="Visual Studio 14 2015 Win64"
 
 :: Build options
-SET BUILD_PYTHON=ON
-SET BUILD_RUNTIME=OFF
+set BUILD_PYTHON=ON
+set BUILD_RUNTIME=OFF
+
+:: Optional libraries
+set USE_CUDA=ON
+set USE_CUDNN=ON
+set USE_OPENMP=ON
+set USE_AVX=ON
+set USE_AVX2=ON
+set USE_FMA=ON
 
 :: Protobuf SDK options
-SET PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
+set PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
 
 :: Protobuf Compiler options
-:: Set the protobuf compiler(i.e., protoc) if necessary
-:: If not, a compiler in the sdk or environment will be used
-SET PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
+:: Set the protobuf compiler(i.e., protoc) if necessary.
+:: If not, a compiler in the sdk or environment will be used.
+set PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
 
 :: Python options
-:: Set your python "interpreter" if necessary
-:: If not, a default interpreter will be used
-:: SET PYTHON_EXECUTABLE=X:/Anaconda3/python
+:: Set your python "interpreter" if necessary.
+:: If not, a default interpreter will be used.
+:: set PYTHON_EXECUTABLE=X:/Anaconda3/python
+
 if %BUILD_PYTHON% == ON (
   if NOT DEFINED PYTHON_EXECUTABLE (
     for /F %%i in ('python -c "import sys;print(sys.executable)"') do (set PYTHON_EXECUTABLE=%%i)
@@ -47,6 +57,12 @@ cmake .. ^
   -G%CMAKE_GENERATOR% ^
   -DBUILD_PYTHON=%BUILD_PYTHON% ^
   -DBUILD_RUNTIME=%BUILD_RUNTIME% ^
+  -USE_CUDA==%USE_CUDA% ^
+  -USE_CUDNN==%USE_CUDNN% ^
+  -USE_OPENMP==%USE_OPENMP% ^
+  -USE_AVX==%USE_AVX% ^
+  -USE_AVX2==%USE_AVX2% ^
+  -USE_FMA==%USE_FMA% ^
   -DTHIRD_PARTY_DIR=%THIRD_PARTY_DIR% ^
   -DPROTOBUF_SDK_ROOT_DIR=%PROTOBUF_SDK_ROOT_DIR% ^
   -DPROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_PROTOC_EXECUTABLE% ^

scripts/build_windows_vc141.bat

-:: #############################################################################
-:: Example command to build on Windows for Visual Studio 2017 (VC141).
-:: #############################################################################
+:: ###############################################################
+:: Command file to build on Windows for Visual Studio 2017 (VC141)
+:: ###############################################################
 
 @echo off
 setlocal
 
-SET ORIGINAL_DIR=%cd%
-SET REPO_ROOT=%~dp0%..
-SET DRAGON_ROOT=%REPO_ROOT%\dragon
-SET THIRD_PARTY_DIR=%REPO_ROOT%\third_party
-SET CMAKE_GENERATOR="Visual Studio 15 2017 Win64"
+:: Build variables
+set ORIGINAL_DIR=%cd%
+set REPO_ROOT=%~dp0%..
+set DRAGON_ROOT=%REPO_ROOT%\dragon
+set THIRD_PARTY_DIR=%REPO_ROOT%\third_party
+set CMAKE_GENERATOR="Visual Studio 15 2017 Win64"
 
 :: Build options
-SET BUILD_PYTHON=ON
-SET BUILD_RUNTIME=OFF
+set BUILD_PYTHON=ON
+set BUILD_RUNTIME=OFF
+
+:: Optional libraries
+set USE_CUDA=ON
+set USE_CUDNN=ON
+set USE_OPENMP=ON
+set USE_AVX=ON
+set USE_AVX2=ON
+set USE_FMA=ON
 
 :: Protobuf SDK options
-SET PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
+set PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
 
 :: Protobuf Compiler options
-:: Set the protobuf compiler(i.e., protoc) if necessary
-:: If not, a compiler in the sdk or environment will be used
-SET PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
+:: Set the protobuf compiler(i.e., protoc) if necessary.
+:: If not, a compiler in the sdk or environment will be used.
+set PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
 
 :: Python options
-:: Set your python "interpreter" if necessary
-:: If not, a default interpreter will be used
-:: SET PYTHON_EXECUTABLE=X:/Anaconda3/python
+:: Set your python "interpreter" if necessary.
+:: If not, a default interpreter will be used.
+:: set PYTHON_EXECUTABLE=X:/Anaconda3/python
+
 if %BUILD_PYTHON% == ON (
   if NOT DEFINED PYTHON_EXECUTABLE (
     for /F %%i in ('python -c "import sys;print(sys.executable)"') do (set PYTHON_EXECUTABLE=%%i)
@@ -47,6 +57,12 @@ cmake .. ^
   -G%CMAKE_GENERATOR% ^
   -DBUILD_PYTHON=%BUILD_PYTHON% ^
   -DBUILD_RUNTIME=%BUILD_RUNTIME% ^
+  -USE_CUDA==%USE_CUDA% ^
+  -USE_CUDNN==%USE_CUDNN% ^
+  -USE_OPENMP==%USE_OPENMP% ^
+  -USE_AVX==%USE_AVX% ^
+  -USE_AVX2==%USE_AVX2% ^
+  -USE_FMA==%USE_FMA% ^
   -DTHIRD_PARTY_DIR=%THIRD_PARTY_DIR% ^
   -DPROTOBUF_SDK_ROOT_DIR=%PROTOBUF_SDK_ROOT_DIR% ^
   -DPROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_PROTOC_EXECUTABLE% ^

scripts/build_windows_vc142.bat

-:: #############################################################################
-:: Example command to build on Windows for Visual Studio 2019 (VC142).
-:: #############################################################################
+:: ###############################################################
+:: Command file to build on Windows for Visual Studio 2019 (VC142)
+:: ###############################################################
 
 @echo off
 setlocal
 
-SET ORIGINAL_DIR=%cd%
-SET REPO_ROOT=%~dp0%..
-SET DRAGON_ROOT=%REPO_ROOT%\dragon
-SET THIRD_PARTY_DIR=%REPO_ROOT%\third_party
-SET CMAKE_GENERATOR="Visual Studio 16 2019"
+:: Build variables
+set ORIGINAL_DIR=%cd%
+set REPO_ROOT=%~dp0%..
+set DRAGON_ROOT=%REPO_ROOT%\dragon
+set THIRD_PARTY_DIR=%REPO_ROOT%\third_party
+set CMAKE_GENERATOR="Visual Studio 16 2019"
 
 :: Build options
-SET BUILD_PYTHON=ON
-SET BUILD_RUNTIME=OFF
+set BUILD_PYTHON=ON
+set BUILD_RUNTIME=OFF
+
+:: Optional libraries
+set USE_CUDA=ON
+set USE_CUDNN=ON
+set USE_OPENMP=ON
+set USE_AVX=ON
+set USE_AVX2=ON
+set USE_FMA=ON
 
 :: Protobuf SDK options
-SET PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
+set PROTOBUF_SDK_ROOT_DIR=%THIRD_PARTY_DIR%\protobuf
 
 :: Protobuf Compiler options
-:: Set the protobuf compiler(i.e., protoc) if necessary
-:: If not, a compiler in the sdk or environment will be used
-SET PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
+:: Set the protobuf compiler(i.e., protoc) if necessary.
+:: If not, a compiler in the sdk or environment will be used.
+set PROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_SDK_ROOT_DIR%\bin\protoc
 
 :: Python options
-:: Set your python "interpreter" if necessary
-:: If not, a default interpreter will be used
-:: SET PYTHON_EXECUTABLE=X:/Anaconda3/python
+:: Set your python "interpreter" if necessary.
+:: If not, a default interpreter will be used.
+:: set PYTHON_EXECUTABLE=X:/Anaconda3/python
+
 if %BUILD_PYTHON% == ON (
   if NOT DEFINED PYTHON_EXECUTABLE (
     for /F %%i in ('python -c "import sys;print(sys.executable)"') do (set PYTHON_EXECUTABLE=%%i)
@@ -48,6 +58,12 @@ cmake .. ^
   -Ax64 ^
   -DBUILD_PYTHON=%BUILD_PYTHON% ^
   -DBUILD_RUNTIME=%BUILD_RUNTIME% ^
+  -USE_CUDA==%USE_CUDA% ^
+  -USE_CUDNN==%USE_CUDNN% ^
+  -USE_OPENMP==%USE_OPENMP% ^
+  -USE_AVX==%USE_AVX% ^
+  -USE_AVX2==%USE_AVX2% ^
+  -USE_FMA==%USE_FMA% ^
   -DTHIRD_PARTY_DIR=%THIRD_PARTY_DIR% ^
   -DPROTOBUF_SDK_ROOT_DIR=%PROTOBUF_SDK_ROOT_DIR% ^
   -DPROTOBUF_PROTOC_EXECUTABLE=%PROTOBUF_PROTOC_EXECUTABLE% ^

tensorflow/core/framework/dtypes.py

@@ -529,7 +529,7 @@ def as_dtype(type_value):
 
     Parameters
     ----------
-    type_value : object
+    type_value : Any
         The data type.
 
     Returns