Open arpgh opened 9 years ago
Looks like we are hitting (1) and (2) described in https://github.com/zigvu/khajuri/issues/87#issuecomment-72309016
With frame density as 1. The steps can be as large as 100 ( since we have 4 video reader processes now )
Since (1) and (2) are not unique to just issue87 i will merge their fixes using this issue.
Fix merged into issue94 branch at 98d5e255c8b2fddfc86a83db3817323a3374ed39. Tested with 14 minute video
Run still getting stuck at end. If this is related to/dependent on #87, can be labled enhancement to be tackled later together.
Not sure, why I don't see this in GPU1.
Waiting for videoDbManagerProcess process to complete.
takes a while but does complete for me in GPU1. Perhaps something specific about GPU2?
Main difference is that with 4 GPUs, gpu2 has 4 frame extraction threads running. (while gpu1 has 2)
Came across this while experimenting with different _framedensity for json comparison. All jsons are produced but the run is stuck at -
Waiting for videoDbManagerProcess process to complete.
Path: gpu2:/mnt/data/wc14cls/training/0seed/vdos/test-issue91/config_01fps