[Feature] Refactor Evaluation for translation models

[plyfager]

Refactor Evaluation for translation models.

[codecov-commenter]

Codecov Report

Merging #126 (8ff8c4a) into master (7e3aacc) will decrease coverage by 0.04%.
The diff coverage is 80.00%.

@@            Coverage Diff             @@
##           master     #126      +/-   ##
==========================================
- Coverage   75.28%   75.23%   -0.05%     
==========================================
  Files         121      121              
  Lines        8189     8190       +1     
  Branches     1561     1560       -1     
==========================================
- Hits         6165     6162       -3     
- Misses       1609     1611       +2     
- Partials      415      417       +2     

Flag	Coverage Δ
unittests	75.23% <80.00%> (-0.05%)	⬇️

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

Impacted Files	Coverage Δ
mmgen/apis/inference.py	48.42% <75.00%> (+0.54%)	⬆️
mmgen/datasets/paired_image_dataset.py	86.04% <100.00%> (ø)
...chitectures/stylegan/generator_discriminator_v1.py	88.70% <0.00%> (-1.13%)	⬇️
mmgen/models/architectures/stylegan/mspie.py	90.59% <0.00%> (-1.00%)	⬇️

---

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 7e3aacc...8ff8c4a. Read the comment docs.

[LeoXing1996]

• Maybe we should support online evaluation in translation_eval.py.
• EvalHook for translation model should be supported as well.

[plyfager]

Of course, the part @LeoXing1996 mentioned is in my plan. I will complete the online evaluation in this PR, and add TranslationEvalHook in a new PR.

[plyfager]

I tested evaluation(include online mode) and inference, and the codes work fine. However, there are some differences between the evaluation results and the previous ones, and I am looking for the reasons.

[nbei]

Please check the file name in this PR and keep them consistent with others in MMGEN. The major problem is that in and the position of the dataset information.

[plyfager]

Well, the file name correction was done in #118 .