Update docs and README.md

Author: Priya Goyal

README.md

@@ -56,7 +56,7 @@ After a few generations of autotuning on a 2-GPU P100 system, we see results res
 We have not yet characterized the precise fraction of peak performance we obtain but it is not uncommon to obtain 80%+ of peak shared memory bandwidth after autotuning. Solid register-level optimizations are still in the work but TC in its current form already addresses the productivity gap between the needs of research and the needs of production. Which is why we are excited to share it with the entire community and bring this collaborative effort in the open.
 
 # Documentation, Environment and Prerequisites
-We provide pre-built VM images in the docker subdirectory, they can be downloaded from dockerhub. We use and support those VMs as part of our continuous integration. Note that we can cross-compile CUDA (but not execute) even if the machine has no physical GPUs. In any case the CUDA toolkit and libraries should always be installed, for now.
+We provide pre-built docker images in the docker subdirectory, they can be downloaded from [dockerhub](https://hub.docker.com/u/tensorcomprehensions/). We use and support those images as part of our continuous integration. Note that we can cross-compile CUDA (but not execute) even if the machine has no physical GPUs. In any case the CUDA toolkit and libraries should always be installed, for now.
 
 To get started, see the [docs](master/docs) directory.
 

docs/Makefile

@@ -6,7 +6,7 @@ SPHINXOPTS    =
 SPHINXBUILD   = sphinx-build
 SPHINXPROJ    = TensorComprehensions
 SOURCEDIR     = source
-BUILDDIR      = build
+BUILDDIR      = ../../TensorComprehensions-docs
 
 # Put it first so that "make" without argument is like "make help".
 help:
@@ -17,4 +17,4 @@ help:
 # Catch-all target: route all unknown targets to Sphinx using the new
 # "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
 %: Makefile
-	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)

docs/source/conf.py

@@ -62,8 +62,8 @@
 
 # General information about the project.
 project = 'Tensor Comprehensions'
-copyright = '2018, Tensor Comprehensions contributors and users'
-author = 'Tensor Comprehensions contributors and users'
+copyright = '2017-present, Facebook, Inc.'
+author = 'Tensor Comprehensions Contributors'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the

docs/source/contacts.rst

@@ -0,0 +1,28 @@
+Contacts
+========
+
+Tensor Comprehensions are in active development and constant improvement.  We welcome your feedback and contributions.  Tensor Comprehensions team is accessible through different channels, choose the one you prefer or the one that seems most appropriate for your communication.
+
+Bugs and features
+-----------------
+
+Found a bug?  Want a feature?  Open an `issue on GitHub <https://github.com/facebookresearch/TensorComprehensions/issues>`_.
+
+
+Mailing list
+------------
+
+Not sure whether the behavior you see is a bug or a feature?  Want to contact the team about something else than code, like an idea for collaboration?  Drop us an email at tensorcomp@fb.com.
+
+Contributions
+-------------
+
+Want to contribute?  Open a `pull request on Github <https://github.com/facebookresearch/TensorComprehensions/pulls>`_.
+
+Don't forget to read the `contributor's instructions <https://github.com/facebookresearch/TensorComprehensions/blob/master/CONTRIBUTING.md>`_.  Different parts of Tensor Comprehensions are managed by different people, make sure to tag the `code owners <https://github.com/facebookresearch/TensorComprehensions/blob/master/CodeOwners.md>`_ of the part you are modifying.
+
+
+Slack channel
+-------------
+
+For a faster and tighter interaction, join our team on Slack: `TensorComprehensions.slack.com <https://tensorcomprehensions.slack.com>`_.  You may need an invitation to join, contact us by email at tensorcomp@fb.com to get one.

docs/source/docker_image.rst

@@ -2,7 +2,8 @@ Installing TC from docker image
 ===============================
 
 We provide docker runtime images for :code:`conda` and :code:`non-conda` environments both. TC officially supports
-running gcc 4.*, CUDA 8, CUDNN 6 and ubuntu14.04 and gcc 5.*, CUDA9, CUDNN6 on ubuntu16.04
+running gcc 4.*, CUDA 8, CUDNN 6 and ubuntu14.04 and gcc 5.*, CUDA9, CUDNN6 on ubuntu16.04. You can find all available images
+for Tensor Comprehensions at the `dockerhub <https://hub.docker.com/u/tensorcomprehensions/>`_
 
 The conda and non-conda images for each setup are below:
 

docs/source/index.rst

@@ -38,3 +38,15 @@ Tensor Comprehensions provides framework-Agnostic Abstractions for High-Performa
    installation_conda_dep
    installation_conda
    installation_non_conda
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Paper
+
+   report
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Support
+
+   contacts

docs/source/ml_with_tc.rst

@@ -79,10 +79,9 @@ TC
 ^^
 
 The current TC implementation sits somewhere here; less verbose than Halide,
-more verbose than matrix algebra. The biggest current negative point is its
-non-intuitive behavior that depends on the inference procedure. But this
-inference procedure can be described properly and follows an intuitive enough
-mental model, see :ref:`inference`.
+more verbose than matrix algebra. The inference procedure has been one subtle 
+tradeoff in TC. It has been designed to follow an intuitive enough mental model,
+but may still evolve in the future towards greater expressiveness, see :ref:`inference`.
 
 Matrix Languages
 ^^^^^^^^^^^^^^^^

docs/source/performance.rst

@@ -2,7 +2,7 @@ Performance of TC
 =================
 
 TC can generate competitive code in a variety of cases thanks to its
-Autotuner (see our companion paper: LINK).
+Autotuner (see our companion paper: `ArXiV <link>`_).
 We will provide a set of benchmarks to illustrate the cases in
 which it is recommended to use TC.
 

docs/source/report.rst

@@ -0,0 +1,17 @@
+Tech Report
+===========
+
+**Authors**:
+
+`Nicolas Vasilache <https://scholar.google.com/citations?user=vIGcvLsAAAAJ&hl=en&oi=ao>`_ (FAIR),
+`Oleksandr Zinenko <https://ozinenko.com>`_ (Inria & DI ENS),
+`Theodoros Theodoridis <theodort@student.ethz.ch>`_ (ETH Zürich),
+`Priya Goyal <https://scholar.google.com/citations?user=-9yiQMsAAAAJ&hl=en>`_ (FAIR),
+`Zachary DeVito <zdevito@fb.com>`_ (FAIR),
+`William S. Moses <http://wsmoses.com>`_ (MIT CSAIL),
+`Sven Verdoolaege <sven@cs.kuleuven.be>`_ (FAIR),
+`Andrew Adams <https://andrew.adams.pub/>`_ (FAIR),
+`Albert Cohen <https://who.rocq.inria.fr/Albert.Cohen>`_ (Inria & DI ENS & FAIR)
+
+We provide more details about Tensor Comprehensions in our tech report which can be found
+here on `ArXiV <link>`_