longyuanli / GRIN_NeurIPS21

GRIN: Generative Relation and Intention Network for Multi-agent Trajectory Prediction

Official Implementation of Generative Relation and Intention Network (GRIN) in PyTorch and DGL.

Dependencies

• torch==1.8.1
• numpy==1.19.2
• scipy==1.6.1
• dgl_cu110==0.6.1
• dgl==0.6.1
• tensorboardX==2.2

Running the code

0. Install all dependencies mentioned above

1. Generate charged dataset for training (NBA dataset is available on [44])

python simulator.py --seed 0 --num_sample 5000 --filename train.npz
python simulator.py --seed 1 --num_sample 1000 --filename test.npz
python simulator.py --seed 2 --num_sample 1000 --filename valid.npz

2. Train the model

bash train.sh

3. Evaluate the model

bash eval.sh