The question about num_features

[wnma3mz]

When Training with FRSKD,

In train, the first feat has been deleted in feat_lst.

FRSKD/classification/utils.py

Lines 60 to 61 in bed2601

	feats = feats[-args.num_features:]
	bi_feats, bi_outputs = bifpn(feats, args.distill == 'od')

But, in model bifpn forward, it still deletes the first feat.

FRSKD/classification/models/bifpn.py

Lines 27 to 28 in bed2601

	def forward(self, feats, preact=False):
	feats = feats[-self.num_features:]

My guess, based on network_channels, is that only one delete operation is required.

FRSKD/classification/main.py

Lines 75 to 76 in bed2601

	args.network_channels = model.network_channels[-args.num_features:]
	bifpn = models.__dict__[args.bifpn](args.network_channels, args.num_classes, args)

So, the following line should also be deleted

FRSKD/classification/models/bifpn.py

Line 28 in bed2601

feats = feats[-self.num_features:]