Issue with testing using distributed and mpii

[pablovela5620]

So I was trying to test the mpii dataset using both test.py and ./tool/dist_test.sh and came across this small issue. It looks like mpii uses the evaluation metric PCKh rather than mPA so I changed the args as follows

For test.py

 python tools/test.py configs/top_down/hrnet/mpii/hrnet_w32_mpii_256x256.py ~/0Dev/data/models/hrnet_w32_mpii_256x256-6c4f923f_20200812.pth --eval PCKh

for ./tools/dist_test.sh

./tools/dist_test.sh configs/top_down/hrnet/mpii/hrnet_w32_mpii_256x256.py ~/0Dev/data/models/hrnet_w32_mpii_256x256-6c4f923f_20200812.pth 2 --eval PCKh

I get the assertion error saying that metric != PCKh and modified the error so I could figure out what was going on and saw

assert metric == 'PCKh', f"got {metric}"
AssertionError: got ['PCKh']

I'm fairly new to this repo so I don't know the ins and outs enough to make a pull request, so I wanted to make you guys aware of this issue!

[jin-s13]

Thanks for reporting. I will check this.

[jin-s13]

Will be fixed in this commit.

[pablovela5620]

I also just noticed that it looks like most mpii configs have their evaluation = dict(interval=1, metric='mPA') rather than 'PCKh` which will cause errors when evaluating

[jin-s13]

Thanks! Yes, It should be PCKh.
We will fix the typos.

[jin-s13]

Fixed in #93