apache--tvm/docs

TVM Documentation

This folder contains the source of TVM's documentation, hosted at https://tvm.apache.org/docs

Build Locally

With Docker (recommended)

1. Build TVM and the docs inside the tlcpack/ci-gpu image using the ci.py script.

   # If this runs into errors, try cleaning your 'build' directory
   python tests/scripts/ci.py docs
   
   # See other doc building options
   python tests/scripts/ci.py docs --help
   

2. Serve the docs and visit http://localhost:8000 in your browser

   # Run an HTTP server you can visit to view the docs in your browser
   python tests/scripts/ci.py serve-docs
   

Native

1. Build TVM first in the repo root folder, then make it importable:

   export TVM_HOME=/path-to-tvm
   export TVM_LIBRARY_PATH=$TVM_HOME/build
   pip install --target=$TVM_HOME/python $TVM_HOME/3rdparty/tvm-ffi
   export PYTHONPATH=$TVM_HOME/python:$PYTHONPATH
   

   docs/conf.py unconditionally imports tvm at startup, so the build will fail immediately if TVM is not importable.

2. Install dependencies

   # Pillow on Ubuntu may require libjpeg-dev from apt
   ./docker/bash.sh ci_gpu -c \
       'python3 -m pip install --quiet sphinx-book-theme==1.1.4 && python3 -m pip freeze' > frozen-requirements.txt
   
   pip install -r frozen-requirements.txt
   

3. Generate the docs

   # TVM_TUTORIAL_EXEC_PATTERN=none skips the tutorial execution to the build
   # work on most environments (e.g. MacOS).
   export TVM_TUTORIAL_EXEC_PATTERN=none
   
   cd docs
   make html
   

4. Run an HTTP server and visit http://localhost:8000 in your browser

   cd _build/html && python3 -m http.server
   

Only Execute Specified Tutorials

The document build process will execute all the tutorials in the sphinx gallery. This will cause failure in some cases when certain machines do not have necessary environment. You can set TVM_TUTORIAL_EXEC_PATTERN to only execute the path that matches the regular expression pattern.

For example, to only build tutorials under /get_started/tutorials, run

python tests/scripts/ci.py docs --tutorial-pattern=/get_started/tutorials

To only build one specific file, do

# The slash \ is used to get . in regular expression
python tests/scripts/ci.py docs --tutorial-pattern=file_name\.py

Helper Scripts

The following script mirrors the CI docs pipeline: it runs a sphinx pre-check (when not running locally) and then performs a full make htmldepoly build, including tutorial execution. You will need a GPU CI environment.

tests/scripts/task_python_docs.sh

To build docs locally without executing tutorials (fastest local iteration):

cd docs && TVM_TUTORIAL_EXEC_PATTERN=none make html

Note: the sphinx pre-check (warning validation) only runs in CI (IS_LOCAL=0). python tests/scripts/ci.py docs always sets IS_LOCAL=1 and skips the pre-check regardless of other flags.

To run the full build including tutorial executions:

python tests/scripts/ci.py docs --full

Define the Order of Tutorials

You can define the order of tutorials with subsection_order and within_subsection_order in conf.py. By default, the tutorials within one subsection are sorted by filename.

Google Colab Integration

All the TVM tutorials can be opened and used interactively in Google Colab by clicking the button at the top of the page. To do this, sphinx-gallery builds .ipynb files from each tutorial, which are automatically deployed to the apache/tvm-site repo's asf-site branch by @tvm-bot.

To make sure your tutorial runs correctly on Colab, any non-Python parts of the tutorial (e.g. dependency installations) should be prefixed by an IPython magic command. These will not be included in the built HTML file. For example, to install Pytorch in your tutorial, add a ReStructured Text block like the following:

######################################################################
# To run this tutorial, we must install PyTorch:
#
# .. code-block:: bash
#
#     %%shell
#     pip install torch
#

Interactive Bash Scripts

In stock IPython, the %%bash magic command should be used to run shell commands. However, this command does not give real-time output - the tutorial's user will not see any output until the entire cell finishes running. When running commands that take several minutes (e.g. installing dependencies), this is annoying.

Luckily, Google Colab has the %%shell magic command that does the same thing as %%bash, but gives output in real time. This command is specific to Colab, and its source code is public. Thus, %%shell should be used instead of %%bash when writing TVM tutorials.