Fix typo in BaseHead default loss_factor.

[SuX97]

No description provided.

[innerlee]

The problem is deeper. There must be some **kwargs not being used and also not being checked

[codecov]

Codecov Report

Merging #446 (f99d07b) into master (5e0ffc1) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #446   +/-   ##
=======================================
  Coverage   86.59%   86.59%           
=======================================
  Files         101      101           
  Lines        7201     7201           
  Branches     1161     1161           
=======================================
  Hits         6236     6236           
  Misses        733      733           
  Partials      232      232           

Flag	Coverage Δ
unittests	86.58% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmaction/models/heads/base.py	86.84% <ø> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5e0ffc1...3a6f1f0. Read the comment docs.

[SuX97]

The problem is deeper. There must be some **kwargs not being used and also not being checked

There is no **kwargs used in building heads and losses. The reason why we haven't tested it and no error being revealed is that, in BaseWeightedLoss, loss_weight is set to be 1.0 by default. And in all our usages(e.g. all configs), we override the loss_cls by dict(type='XXXLoss') without specifying loss_weight.

[innerlee]

Is loss_factor a valid argument?

[SuX97]

Is loss_factor a valid argument?

Yes, but it will always be overridden together with the total dictionary of loss_cls. So we never met this error before.