2812 enhance swish and efficientnet

[yiheng-wang-nv]

Signed-off-by: Yiheng Wang [email protected]

Fixes #2812 .

This PR also enhances efficientnet via:

1. adding EfficientNetBNFeatures class, which is used to build a backbone for segmentation and object detection tasks.
2. remove unnecessary code (load fc) for _load_state_dict, and remove the hard requirement of in_channels = 3. Now, the network with in_channels != 3 will load weights for all layers except the first conv layer.
3. add more weights urls.

The enhancement refers to: https://github.com/rwightman/pytorch-image-models

Status

Ready

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• Quick tests passed locally by running ./runtests.sh --quick --unittests.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

[wyli]

/build

[wyli]

/build

[wyli]

/build

[wyli]

Thanks, there are some minor style issues reported here https://github.com/Project-MONAI/MONAI/pull/2813/checks?check_run_id=3419685439 please help fix them here or in another PR

[wyli]

/build

[wyli]

/build