Several events in sample

[hushchyn-mikhail]

Could you please add to the createDataset() function possibility to generate several events and save it into one file?

[dhrou]

Running today 2017/02/15 the code as:
python2.7 build_datasets.py --output-dir ../data/ --num-events 3 --hits-per-event 10000

Several events are generated (I believe) from this hits file:
head -10 ../data/dataset_trackml/hits.csv
3,35100,-4998.57153668,-3318.82974827
1,48149,-83.4828668782,6999.50216879
1,71709,-6562.79287918,2435.10772348
3,3224,-2861.94932439,-899.657323761
1,49170,-2848.73767301,940.581762459
3,10524,-2228.09978463,7683.46089661
1,23489,-5913.88500405,1012.89888878
2,36475,-5781.67959668,5529.2440268
1,71784,4814.89545659,1347.8960663
1,68426,2949.8993369,-5224.74449428

However the hits from the different events are mixed up. This should no the the case.

[tboser]

The reason they are mixed up is because I shuffled them randomly, I can fix that and have hits and events stay together (you could also sort the file by the first row if you'd like a quicker result). As for the additional problem with 1 hit per event and per track that simply shows that there is 1 hit with ID '123' in each event. I can make the IDs not overlap for different events if you'd like?

[dhrou]

I have a fix for this, I'll do this, as my first commit (I hope you did not do it).
(for the second thing, this was a stupid misunderstanding of mine, it is perfectly fine to have hit barcode reused in different events)