ResNeSt-MegEngine

The MegEngine Implementation of ResNeSt.

Usage

Install dependency.

pip install -r requirements.txt

Convert trained weights from torch to megengine, the converted weights will be save in ./pretained/

python convert_weights.py -m resnest50

Import from megengine.hub:

Way 1:

from  megengine import hub
modelhub = hub.import_module(repo_info='asthestarsfalll/resnest-megengine', git_host='github.com')

# load ResNeSt model and custom on you own
resnest = modelhub.ResNet(modelhub.Bottleneck, [3, 24, 36, 3], radix=2, groups=1, bottleneck_width=64,
                          deep_stem=True, stem_width=64, avg_down=True, avd=True, avd_first=False)

# load pretrained model 
pretrained_model = modelhub.resnest50(pretrained=True) 

Way 2:

from  megengine import hub

# load pretrained model 
model_name = 'resnest50'
pretrained_model = hub.load(
    repo_info='asthestarsfalll/resnest-megengine', entry=model_name, git_host='github.com', pretrained=True)

Currently only support resnest50, you can run convert_weights.py to convert other models. For example:

  python convert_weights.py -m resnest101

Then load state dict manually.

model = modelhub.convnext_small()
model.load_state_dict(mge.load('./pretrained/resnest101.pkl'))
# or
model_name = 'resnest101'
model = hub.load(
    repo_info='asthestarsfalll/resnest-megengine', entry=model_name, git_host='github.com')
model.load_state_dict(mge.load('./pretrained/resnest101.pkl'))

TODO

• add train codes
• maybe export to some inference framwork

Reference

The official implementation of ResNeSt