If the pytorch source code has been already build with ninja, and there’s a very small change, such as printf, in one c++ code, not headers. what’s the most convenient way to rebuild it? What’s your best practices.
The CONTRIBUTING docs mention some useful workflows, such as using
ccache etc. for incremental builds, which can speedup the development significantly.
However, as you commented in other posts, after we modify some
cpp files, running again
python setup.py develop is enough to rebuild the changes.
Is the guide in CONTRIBUTING out-dated?