fix basemodule

[jshilong]

Motivation

In the original implementation, when init a module with the pre-train model,

mmcv/mmcv/runner/base_module.py

Line 44 in c77e95a

for m in self.children():

if there are children who have init_weights, the pretrain model would be overridden which is not reasonable.

Modification

Retune when find it is loading a pre-train model.

BC-breaking (Optional)

None

[MeowZheng]

Many thanks for you fixing the bug. It looks good to me.

[codecov]

Codecov Report

Merging #1018 (97c5ccc) into master (c77e95a) will increase coverage by 0.14%.
The diff coverage is 50.00%.

@@            Coverage Diff             @@
##           master    #1018      +/-   ##
==========================================
+ Coverage   64.70%   64.84%   +0.14%     
==========================================
  Files         152      154       +2     
  Lines        9841     9886      +45     
  Branches     1791     1800       +9     
==========================================
+ Hits         6368     6411      +43     
- Misses       3144     3145       +1     
- Partials      329      330       +1     

Flag	Coverage Δ
unittests	64.84% <50.00%> (+0.14%)	⬆️

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

Impacted Files	Coverage Δ
mmcv/runner/base_module.py	75.67% <50.00%> (-3.12%)	⬇️
mmcv/ops/__init__.py	100.00% <0.00%> (ø)
mmcv/ops/pixel_group.py	100.00% <0.00%> (ø)
mmcv/ops/contour_expand.py	100.00% <0.00%> (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 c77e95a...97c5ccc. Read the comment docs.