npu.distribute.npu_distributed_keras_optimizer_wrapper

Applicability

Product	Supported
Atlas 350 Accelerator Card	√
Atlas A3 training product/Atlas A3 inference product	√
Atlas A2 training product/Atlas A2 inference product	√
Atlas 200I/500 A2 inference product	☓
Atlas inference product	☓
Atlas training product	√

Description

Adds the AllReduce operation of the NPU to aggregate the gradients, and then updates the gradients. This API applies only to distributed training.

Prototype

def npu_distributed_keras_optimizer_wrapper(optimizer, reduce_reduction="mean", fusion=1, fusion_id=-1, group="hccl_world_group")

Parameters

Parameter	Input/Output	Description
optimizer	Input	TensorFlow gradient training optimizer.
reduce_reduction	Input	A string. Aggregation operation type. The value can be mean, max, min, prod, or sum.
fusion	Input	An int. AllReduce operator fusion flag. The value can be one of the following: 0: The AllReduce operator is not fused with other AllReduce operators during network compilation. 1: The AllReduce operator is fused based on the gradient splitting policy during network compilation. 2: AllReduce operators with the same fusion_id are fused during network compilation.
fusion_id	Input	An int. AllReduce operator fusion ID. If fusion is set to 2, AllReduce operators with the same fusion_id are fused during network compilation.
group	Input	A string of up to 128 bytes, including the end character. Group name, which can be a user-defined value or hccl_world_group.

Returns

TensorFlow gradient training optimizer.

Example

import npu_device as npu
optimizer = tf.keras.optimizers.SGD()
optimizer = npu.distribute.npu_distributed_keras_optimizer_wrapper(optimizer) # Use NPU-based distributed computing to update gradients.
model.compile(optimizer = optimizer)

Parent topic: TF Adapter 2.x API Reference