Defines the automatic batching configuration. More...

Detailed Description

Defines the automatic batching configuration.

Note: With the automatic batching method, the number of batches on rays and on rx channels are computed by the system taking into account the specified two constraints. There are two possibilities for the second constraint according to your needs: To use the default GPU, set the gpu_memory_quota. To use any other GPU or several GPUs, set the gpu_quotas.

Member Data Documentation

int32 vss::simulation::AutomaticBatching::max_number_of_ray_batches = 1

First constraint: the upper bound on the automatically computed number of ray batches.

double vss::simulation::AutomaticBatching::gpu_memory_quota = 2

Second constraint, option 1: the max fraction (quota) of the total GPU memory that can be used. The default GPU (whose index is 0) will be used.

map<string, double> vss::simulation::AutomaticBatching::gpu_quotas = 3

Second constraint, option 2: the max fraction (quota) of the total memory of the specified GPU that can be used.

Key [string]: The GPU name.

Value [double]: The max fraction of the GPU memory that can be used.

Note: The GPU names must correspond to the GpuIdentifier names specified in the deploy parameters.

The documentation for this struct was generated from the following file: