Open Samuel-Fipps opened 1 year ago
@Samuel-Fipps, apologies for the late reply.
Just avoid generating data in test rooms. splits/oct21.json
is specific to the ALFRED dataset. If you use generate_trajectories.py
to generate another dataset you will get a new train and val distribution different from the ALFRED dataset.
Hello,
I'm just generating more data with generate_trajectories.py, I was just wondering if the splits/oct21.json file needed to be changed in anyway to include this new data.
Sorry if this is mentioned somewhere else, I tired to look around for the answer the best I could.