Skip to content

Latest commit

 

History

History
12 lines (9 loc) · 1.53 KB

File metadata and controls

12 lines (9 loc) · 1.53 KB

FAQ

Q: Preprocessing (cache) takes forever

A: Preprocessing could be bottlenecked by file I/O. Make sure your dataset is placed at a location with enough I/O bandwith, eg. local SSDs. Since our feature builder queries the database files, lacking of I/O bandwith bottlenecks the query speed. Also please check the amount of samples you want to preprocess (details in the next point).

Q: How many samples should I use? / Dataset size / Subsampling

A: Preprocessing the full dataset exaustively may not be ideal. By default we use NuPlanScenario so that it generates a sample per timestep in the database, which is 20Hz -- very dense. By setting scenario_filter.limit_total_scenarios=0.1 for example, it reduces to 2Hz so that the number of samples by 1/10. Note that this rate could be independent to the frequency used in input/target features. You can always have your desired sampling rate of input/target features in your customized FeatureBuilder or TargetBuilder when extracting features.

Q: My gradient becomes NaN / Numerical stability / Float precision

A: FP16 causes numerical instability problems. Before deep diving into your own model, make sure your lightning.trainer.params.precision is set to 32. In the default configuration we set lightning.trainer.params.precision=16, meaning that FP16 is used for training. If you are not 100% sure about your model is numerically stable using FP16, please always use FP32.