Distributed Inference on NVIDIA GPUs using TensorFlow 2.18

[obriensystems]

13900A RTX-A4000

https://www.tensorflow.org/guide/distributed_training#multiworkermirroredstrategy

# CommunicationImplementation.RING is RPC-based and supports both CPUs and GPUs.

os.environ["TF_CONFIG"] = json.dumps({
    "cluster": {
        "worker": ["192.168.0.129:4001", "192.168.0.131:4002"]#,
        #"ps": ["192.168.0.129:4003"]
    },
   "task": {"type": "worker", "index": 1}
})

communication_options = tf.distribute.experimental.CommunicationOptions(
    implementation=tf.distribute.experimental.CommunicationImplementation.NCCL)
strategy = tf.distribute.MultiWorkerMirroredStrategy(
    communication_options=communication_options)