mindspore.device_context.ascend.op_tuning.aclnn_cache

View Source On AtomGit
mindspore.device_context.ascend.op_tuning.aclnn_cache(enable_global_cache=False, cache_queue_length=10000)[source]

Configure aclnn cache parameters.

Parameters

  • enable_global_cache (bool, optional) – Whether to set the aclnn cache to global when in GRAPH_MODE. Default: False.

  • cache_queue_length (int, optional) – The cache queue length to set. Default: 10000.

Examples

>>> import mindspore as ms
>>> ms.device_context.ascend.op_tuning.aclnn_cache(True, 10000)