|
Previously the final testing guide mentioned p2.xlarge as the testing platform for GPU-based solutions, for verifying both the training and the inference processes. In case of training verification this is now changed to the following system: 16 Core Intel @ 3.50GHz with 265 GB Ram and 4x Titan Xp Gpu The allowed training time is changed to 8 days for all 4 cities, previously it was 8 days per city.
We are also considering improving the testing server for inference verification, hopefully this will be decided soon.
For CPU-based solutions the setup is not changed, you can assume an m4.10xlarge AWS instance. |