--cpus

CPU indices

Defines the explicit CPUs which NuPIC Inference Server will run on. If not specified, NuPIC Inference Server runs on all available CPU cores.

This can be useful when running NuPIC Inference Server concurrently with other workloads on the same machine.

--expose

None

Expose NuPIC Inference Server's HTTP port (8000), GRPC" port (8001), and metrics port (8002) to all hosts. If not specified, all these ports are exposed on the loopback interface.

This is required to allow remote clients to make requests to the Inference Server.

--gpus

None

Allow NuPIC Inference Server to use all GPU devices. If not specified, NuPIC Inference Server runs with CPU only.

Note that each model will still have to be individual configured to use the GPU using [instance_group.kind](doc:configuration#instance-group) on the respective config.pbtxt.

--log-level

Integer

Set custom log verbosity for NuPIC Inference Server. The level is an integer value. 0 means verbose logging is disabled. 1 and 2 provides different level of verbose loggings. Level 2 enables most detailed logging. If not specified, the default value is 0.

--shm-size

Bytes

Set custom shared memory size (/dev/shm) for NuPIC Inference Server. If not specified, the default shared memory size is 512 MB (512m).

--enable-caching

None

Enable response caching in NuPIC Inference Server. If not specified, caching is disabled.

--cache-size

Bytes

Set custom cache size for NuPIC Inference Server. If not specified, the default cache size is 1GB.

--additional_args

Various

Any additional arguments to pass to Triton Server; e.g., NUMA node configurations.