[Fix] Fix a bug when adding ('val', 1) to workflow

[yezhen17]

Val in workflow is intended to show the losses on test set. This pr fixes a bug when adding val to workflow.

Still, val uses train-time data augmentations, which is not preferred. However, one can slightly modify the config files and train.py in order to use test-time data augmentations for val. Therefore we do not address the issue in this pr. Perhaps some refactors might be considered in the future regarding this issue (the test_mode meaning and pipeline overriding here are also a bit confusing)?

[codecov]

Codecov Report

Merging #370 (8e06812) into master (f0ba0ce) will not change coverage.
The diff coverage is 33.33%.

❗ Current head 8e06812 differs from pull request most recent head 4880ee6. Consider uploading reports for the commit 4880ee6 to get more accurate results

@@           Coverage Diff           @@
##           master     #370   +/-   ##
=======================================
  Coverage   49.55%   49.55%           
=======================================
  Files         177      177           
  Lines       11834    11834           
  Branches     1844     1844           
=======================================
  Hits         5864     5864           
  Misses       5610     5610           
  Partials      360      360           

Flag	Coverage Δ
unittests	49.55% <33.33%> (ø)

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

Impacted Files	Coverage Δ
mmdet3d/ops/knn/knn.py	32.14% <0.00%> (ø)
mmdet3d/__init__.py	77.27% <100.00%> (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 f0ba0ce...4880ee6. Read the comment docs.