Let's use the public trajectory benchmark dataset for easier comparision with baselines. This ETH/UCY trajectory data is one of the most widely used dataset.
You can check the data loading process in 1-eth-dataset-info.ipynb.
I took the baselines value from the Trajectron++ paper for the ETH dataset. Later we could also do it for other four datasets.
Check the baselines.py for the comparison calculation function and 2-eth-compare-with-baselines.ipynb for the comparison with baselines.
I temporarily uploaded the gpt.py for your reference as this is a private repo.
Use the following command to train:
python main.py problem=trajectory_prediction init_pop_size=4 pop_size=4 max_fe=20 timeout=20 llm_client=azureFor CVM, see baselines/cvm/README.md.
TODO