FAQs

• Q: My TensorFlow does not expect the model format.

  A: To use the library you will need specific versions of TensorFlow, please refer to Getting Started

• Q: What Python version should I have?

  A: The bare minimum you should have is Python 3. We tested and recommend using Python 3.6 as higher Python versions may cause problems with certain dependencies.

• Q: The script runs very slow on my machine. What should I do?

  A: First you need to isolate the bottleneck and the cause of slowness. If it’s from data preprocessing, consider parallel processing and follow guide. If it’s the predicting/learning part, then you should try to enable GPU (if you have one) support by installing the GPU-version of tensorflow, or consider using a more powerful machine or a machine with GPU.

• Q: The data is huge; storing and moving it takes a lot space and time.

  A: We recommend enable data compression following guide.

• Q: My job was running on Kubernetes cluster and got killed with status OOMKilled during CHAP pre_process_data.py stage.

  A: One potential reasons could be extremely large ActiGraph 30Hz Raw file, such as > 4GB. Try to remove the large file and re-run.