not predicting trajectory for every agent; validation data

[ahsan155]

I trained the model and have checkpoints. I make prediction by running test.py in tools folder. I checked the config file and it looks like prediction is made on validation data. Running test.py creates output folder which contains result.pkl file. But the issue I have is, this result.pkl file does not contain predicted trajectory for all the agents that are in validation data for a particular scenario. In one scenario of validation data, I have 83 agents but for the same scenario, in result.pkl file there are only 3 agents with trajectory. How can I get trajectory for all agents?

[FQYQC]

Have you pulled the newest commit? Previous codes seem to have this issue, but the newest commit changes the data structure of result.pkl, which should have solve this.

[BJHYZJ]

I also encountered this problem and it seems still not been solved.

[ahsan155]

I figured it out. At the end of tfrecord file, there is field called tracks_to_predict which specifies the index of tracks to make prediction on. In my case, tracks_to_predict field had index of only 3 agents. I inserted more index in this field to do more predictions.