Nuplan数据集的训练速度慢，GPU无作用

在跑tuplan的训练模型时，发现第一次运行时，训练速度很慢，查看了一下nvidia-smi，发现GPU没有工作。这种情况是因为没有生成cache，所以在epoch0时运行速度很慢。

正确的方法是在配置文件中更改

py_func=cache，

cache.force_feature_computation=True

以下是tutorial原文：

• If preprocessing takes a large amount of time, it can cause training to fail (especially in a distributed setting). It may be beneficial to create the feature cache by first running the caching with the argument py_func=cache cache.force_feature_computation=True. This will generate the features using CPU, which should speed up training dramatically. Once caching is complete, supply the overrides cache.force_feature_computation=False cache.cache_path=/path/to/cache cache.use_cache_without_dataset=True to avoid re-computing the features.