CM_WORKER_SIZE

Description

In the TensorFlow distributed training or inference scenario, you can choose not to use the ranktable file. Instead, you can use the environment variables CM_CHIEF_IP, CM_CHIEF_PORT, CM_CHIEF_DEVICE, CM_WORKER_SIZE, and CM_WORKER_IP to automatically generate resource information and initialize the collective communication component.

Configures the number of devices in the service communicator.

The value of this environment variable must be an integer ranging from 0 to 32768.

Example

export CM_WORKER_SIZE=8

Restrictions

This environment variable cannot be used together with RANK_TABLE_FILE, RANK_ID, or RANK_SIZE.

Applicability

Atlas Training Series Product