Train&Training evaluate

[Vio1etovo]

I found that num_workers=32 and default epochs=100 in DataLoder during training. Are these all necessary? Using these default settings, there are many problems in training, such as out of memory, [Errno111] and so on. I reduced these two parameters, and it looks normal at the moment.
In addition, are there any training evaluation function in the code?

[pjl1995]

1. It is not necessary and you can reduce it.
2. In training you can get the .txt files, you can evaluate it by the devkit in MOT challenge: https://github.com/dendorferpatrick/MOTChallengeEvalKit

[fadilkamal]

You can also create a code for evaluation using motmetrics using the .txt generated: https://github.com/cheind/py-motmetrics