PR for addressing issue https://github.com/NVIDIA/TensorRT-Incubator/issues/302. Supporting mixed types is important, since tripy does not have modules like Identity or AveragePool. With this, users can make sequential nets or module lists that include lambda functions or other callables like tp.avgpool.
[x] modified _iterate_members_of_type to have a more relaxed check on checking module/paramater types.
[x] Added mixed container tests to tests/frontend/module/test_module.py
[x] Added mixed container tests to tests/frontend/module/test_sequential.py
[x] Added example mixed container usage to tp.Module
[x] Added example mixed container usage to tp.Sequential
PR for addressing issue https://github.com/NVIDIA/TensorRT-Incubator/issues/302. Supporting mixed types is important, since tripy does not have modules like
Identity
orAveragePool
. With this, users can make sequential nets or module lists that include lambda functions or other callables liketp.avgpool
._iterate_members_of_type
to have a more relaxed check on checking module/paramater types.tests/frontend/module/test_module.py
tests/frontend/module/test_sequential.py
tp.Module
tp.Sequential