MGNet pretraining goes wrong

Hi Yinyu:

I tried to pretrain MGNet with `python main.py configs/mgnet.yaml --mode train` and test it with `python main.py configs/mgnet.yaml --mode test`.

However, after 50 epochs of training, the learning rate quickly reduced to a seemingly unreasonable level of 1e-08 with the best chamfer_loss stuck at 5.67 after the 6th epoch.
[log.txt](https://github.com/yinyunie/Total3DUnderstanding/files/5272828/log.txt)

Also, the test results of the best checkpoint looks like below:
[log.txt](https://github.com/yinyunie/Total3DUnderstanding/files/5272830/log.txt)

Is there anything I missed?

Reproduce the commands `python main.py configs/mgnet.yaml --mode train` and `python main.py configs/mgnet.yaml --mode test`, then inspect `configs/mgnet.yaml` and both linked `log.txt` files. Trace the learning-rate reduction, stalled chamfer loss, and best-checkpoint test output; done means the training behavior has an identified cause and the resulting checkpoint produces validated test results.

Written by the indexing model from the issue text.