[Docs] Add and update documentation (#48)

* .

* update docs

* format & lint

docs/source/_static/custom.css

@@ -55,7 +55,7 @@ code {
     font-family: monospace;
 }
 code {
-    color: #3e3ead;
+    color: #434552;
 }
 
 .heading-style, h1, h2, h3, h4, h5, h6 {

docs/source/conf.py

@@ -88,6 +88,9 @@
 todo_include_todos = True
 
 autodoc_typehints = 'description'
+autodoc_default_options = {
+    'member-order': 'bysource',
+}
 
 bibtex_default_style = 'unsrt'
 bibtex_bibfiles = ['references.bib']

docs/source/getting-started/build-from-source.rst

@@ -0,0 +1,60 @@
+Build from source
+-------------------
+.. _Build-from-source:
+
+If you want to contribute to Hidet, or you encountered any problem installing hidet via pip, it is better to install
+hidet from source.
+
+Clone the code
+~~~~~~~~~~~~~~
+
+First clone the repository to local:
+
+.. code-block:: console
+
+  $ git clone https://github.com/hidet-org/hidet
+
+Build shared libraries
+~~~~~~~~~~~~~~~~~~~~~~
+
+The runtime library is written in C++ and compiled into a shared library. To build the shared library, you need to have
+a C++ compiler installed (as well as build tools like ``cmake``, and ``make``). The following command will build the
+shared library:
+
+.. code-block:: console
+
+  $ cd hidet
+  $ mkdir build
+  $ cd build
+  $ cp ../config.cmake .  # copy the cmake config to build directory
+  $ cmake ..
+  $ make -j4
+
+After building, you could find two libraries ``libhidet.so`` and ``libhidet_runtime.so`` under ``build/lib`` directory.
+
+Update environment variables
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+To allow Python interpreter to find hidet package under ``python`` directory of the repository, we should append the
+directory to ``PYTHONPATH`` variable. To allow the system find the shared libraries we built in the previous step,
+we should append ``build/lib`` directory to ``LD_LIBRARY_PATH`` variable.
+
+.. code-block:: console
+
+  $ export HIDET_HOME=<The Path to Hidet Repo>
+  $ export PYTHONPATH=$PYTHONPATH:$HIDET_HOME/python
+  $ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HIDET_HOME/build/lib
+
+To avoid repeating above commands, it is recommended to put above commands to your shell's initialization script
+(e.g., ``~/.bashrc`` for Bash and ``~/.zshrc`` for Zsh).
+
+Validation
+~~~~~~~~~~
+
+To make sure we have successfully installed hidet, run the following command in a new shell:
+
+.. code-block:: console
+
+  $ python -c "import hidet"
+
+If no error reports, then hidet has been successfully installed on your computer.

docs/source/getting-started/install.rst

@@ -1,89 +1,21 @@
 Installation
 ============
 
-Install via pip
-----------------
-
-.. todo::
-  :class: margin
-
-  Coming soon.
-
-Run the following command to install ``hidet`` package directly using python pip:
+Run the following command to install ``hidet`` package via python pip:
 
 .. code-block:: console
 
   $ pip install hidet
 
-
-Install from source
--------------------
-
-.. note::
-  :class: margin
-
-  Currently, hidet is not open-sourced yet. Please ask Yaoyao get access to the GitHub repository.
-
-  In the future, we will publish hidet to The `Python Package Index (PyPI) <https://pypi.org/>`_.
-
-If you are a developer of Hidet, it is better to install hidet from source.
-
-Clone the code
-~~~~~~~~~~~~~~
-
-First clone the repository to local:
-
-.. code-block:: console
-
-  $ git clone https://github.com/hidet-org/hidet
-
-Build shared libraries
-~~~~~~~~~~~~~~~~~~~~~~
-
-Create the directory to build the shared libraries:
-
-.. code-block:: console
-
-  $ cd hidet
-  $ mkdir build
-  $ cd build
-  $ cp ../config.cmake .  # copy the cmake config to build directory
-
-You could customize the build by update the ``config.cmake`` file, according the instructions in the default config file. Usually,
-just leave it as default.
-
-Build hidet shared libraries (under ``build`` directory):
-
-.. code-block:: console
-
-  $ cmake ..
-  $ make -j4
-
-After building, you could find two libraries ``libhidet.so`` and ``libhidet_runtime.so`` under ``build/lib`` directory.
-
-Update environment variables
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-To allow Python interpreter to find hidet package under ``python`` directory of the repository, we should append the directory to ``PYTHONPATH`` variable.
-To allow the system find the shared libraries we built in the previous step, we should append ``build/lib`` directory to ``LD_LIBRARY_PATH`` variable.
-
-.. code-block:: console
-
-  $ export HIDET_HOME=<The Path to Hidet Repo>
-  $ export PYTHONPATH=$PYTHONPATH:$HIDET_HOME/python
-  $ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HIDET_HOME/build/lib
-
-To avoid repeating above commands, it is recommended to put above commands to your shell's initialization script (e.g., ``~/.bashrc`` for Bash and ``~/.zshrc`` for Zsh).
-
-Validation
-~~~~~~~~~~
-
-To make sure we have successfully installed hidet, run the following command in a new shell:
+To verify the installation, run the following command:
 
 .. code-block:: console
 
   $ python -c "import hidet"
 
-If no error reports, then hidet has been successfully installed on your computer.
+If you want, you can also :doc:`build from source </getting-started/build-from-source>`.
 
+.. toctree::
+   :hidden:
 
+   build-from-source

docs/source/index.rst

@@ -6,15 +6,12 @@
 Welcome to Hidet's Documentation
 ================================
 
-.. note::
-
-  Hidet has not been open-sourced yet. Please do not distribute this website.
-
 Hidet is an open-source DNN inference framework, it features
 
-- **Ease of Use**: Support end to end inference given a model in ONNX format.
+- **Ease of Use**: Support end to end inference for PyTorch and ONNX models.
 - **High Performance**: Graph-level optimizations and operator-level kernel tuning.
-- **Extensibility**: Easy to add new operators and new hardware backends.
+- **Extensibility**: Easy to add new operators, and fusion patterns.
+- **Python Oriented**: All core components are written in Python.
 
 
 .. toctree::

docs/source/python_api/graph/frontend/index.rst

@@ -0,0 +1,8 @@
+hidet.graph.frontend
+====================
+
+.. toctree::
+  :caption: Submodules
+
+  onnx
+  torch

docs/source/python_api/graph/frontend/onnx.rst

@@ -1,6 +1,5 @@
-hidet.graph.frontend
---------------------
+hidet.graph.frontend.onnx
+-------------------------
+
+.. autofunction:: hidet.graph.frontend.from_onnx
 
-.. automodule:: hidet.graph.frontend.onnx
-  :members:
-  :autosummary:

docs/source/python_api/graph/frontend/torch.rst

@@ -0,0 +1,7 @@
+hidet.graph.frontend.torch
+--------------------------
+
+.. autofunction:: hidet.graph.frontend.from_torch
+
+.. autoclass:: hidet.graph.frontend.torch.DynamoConfig
+  :members:

docs/source/python_api/graph/index.rst

@@ -4,13 +4,13 @@ hidet.graph
 .. toctree::
   :caption: Submodules
 
-  frontend/onnx
-  ops/index
+  frontend/index
   transforms/index
 
 
 .. automodule:: hidet.graph
   :members:
   :imported-members:
   :special-members: __call__
+  :exclude-members: Tensor
   :autosummary:

docs/source/python_api/index.rst

@@ -1,6 +1,11 @@
 Python API
 ==========
 
+.. note::
+
+  We are actively working on adding more api documentation.
+
+
 .. toctree::
    :maxdepth: 1
    :caption: Python API
@@ -9,6 +14,8 @@ Python API
    option
    driver
    cuda
+   tensor
+   ops/index
    ir/index
    graph/index
    runtime/index

docs/source/python_api/graph/ops/index.rst → docs/source/python_api/ops/index.rst

@@ -1,11 +1,11 @@
-hidet.graph.ops
-===============
+hidet.ops
+=========
 
 .. todo::
 
   We are still working on the documentation of operators.
 
-.. automodule:: hidet.graph.ops
+.. automodule:: hidet.ops
   :members:
   :undoc-members:
   :imported-members:

docs/source/python_api/tensor.rst

@@ -0,0 +1,38 @@
+hidet.Tensor
+============
+
+
+.. autoclass:: hidet.Tensor
+
+    .. autoattribute:: shape
+    .. autoattribute:: dtype
+    .. autoattribute:: device
+    .. autoattribute:: size
+    .. autoattribute:: nbytes
+    .. autoattribute:: storage
+    .. autoattribute:: trace
+    .. autoattribute:: op
+    .. autoattribute:: layout
+    .. automethod:: tolist
+    .. automethod:: to_device
+    .. automethod:: astype
+    .. automethod:: cpu
+    .. automethod:: cuda
+    .. automethod:: copy
+    .. automethod:: cpu_async
+    .. automethod:: cuda_async
+    .. automethod:: copy_async
+    .. automethod:: detach
+    .. automethod:: numpy
+    .. automethod:: torch
+    .. automethod:: to
+    .. automethod:: item
+    .. automethod:: signature
+    .. automethod:: is_symbolic
+    .. automethod:: contiguous
+    .. automethod:: reshape
+    .. automethod:: squeeze
+    .. automethod:: unsqueeze
+    .. automethod:: rearrange
+    .. automethod:: sum
+    .. automethod:: mean