model baselines

[chan-yuu]

I used the same method:

python navsim/planning/script/run_training.py \
    agent=ego_status_mlp_agent \
    experiment_name=ego_mlp_training \
    train_test_split=navtrain \
    trainer.params.max_epochs=100 \
    cache_path=$NAVSIM_EXP_ROOT/training_cache/ 

python navsim/planning/script/run_training.py \
    agent=transfuser_agent \
    experiment_name=transfuser_training \
    train_test_split=navtrain \
    trainer.params.max_epochs=100 \
    cache_path=$NAVSIM_EXP_ROOT/training_cache/ \

The trained model is evaluated with run_pdm_score, but my score is lower:

transfuser:0.5702315411492036(baseline model:
0.8338995365634392)

ego_mlp:0.5818672986516937(baseline model:0.65)

What is this? It's significantly lower than baseline. I also want to know where the agent of ltf is defined, and look forward to your reply