search

Thinklab-SJTU / Bench2Drive

[NeurIPS 2024 Datasets and Benchmarks Track] Closed-Loop E2E-AD Benchmark Enhanced by World Model RL Expert
Other
1.29k stars 85 forks source link

Scaling Law Curve #95

Closed YangS03 closed 2 months ago

YangS03 commented 2 months ago

I am wondering if you test the scaling law curve when you increase the training dataset since you have released different datasets with different sizes. You choose 1K routes for full-setting and claim that it's comparable to Nuscenes. But Nuscenes is not for training an agent but was developed for an autonomous driving perception system. Could you be so kind as to try to draw the curve using different amounts of data since it consumes a lot of resources to reproduce the result independently?

jiaxiaosong1002 commented 2 months ago

@YangS03 You may do it by yourself.