[Fix] Change the path to pre-trained EDVR-M

[ckkelvinchan]

Motivation

As discussed in #391, there are additional .modules's in edvrm_x4_8x4_600k_reds_20200622-ba4a43e4.pth. Therefore, the weights cannot be correctly loaded.

Modification

We manually removed the unnecessary .modules. The path to the modified model is changed.

[innerlee]

Any idea on the cause?

[codecov]

Codecov Report

Merging #396 (7346dd5) into master (2efae2b) will increase coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #396      +/-   ##
==========================================
+ Coverage   80.43%   80.45%   +0.02%     
==========================================
  Files         186      186              
  Lines        9936     9936              
  Branches     1450     1450              
==========================================
+ Hits         7992     7994       +2     
+ Misses       1749     1748       -1     
+ Partials      195      194       -1     

Flag	Coverage Δ
unittests	80.43% <ø> (+0.01%)	⬆️

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

Impacted Files	Coverage Δ
...ls/components/stylegan2/generator_discriminator.py	86.14% <0.00%> (+1.20%)	⬆️

---

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 2efae2b...7346dd5. Read the comment docs.

[ckkelvinchan]

Any idea on the cause?

No clue at the moment. It seems not the problem of publish_model. I guess model = model.module is not execute for some unknown reasons.

[innerlee]

Would you mind digging into this problem?

[ckkelvinchan]

Would you mind digging into this problem?

Sure I will try

[ckkelvinchan]

I tried training the model and there is no error in this current version.

I checked the commit history and found that two EDVR-M models are uploaded in the same commit. It is kind of strange that only one has this problem. May need more time but not sure whether I can figure it out.