mindspore.device_context

The device_context encapsulates the interface for querying the number of devices with whether the currently specified backend is available.

Cpu Device Backend Management

`mindspore.device_context.cpu.device_count`	Returns the compute-capable device count of CPU.
`mindspore.device_context.cpu.is_available`	Returns whether the CPU backend is available.
`mindspore.device_context.cpu.op_tuning.threads_num`	Set the number of threads for CPU kernel used.

`mindspore.device_context.gpu.device_count`	Returns the number of GPUs available.
`mindspore.device_context.gpu.is_available`	Returns a bool indicating if CUDA is currently available.
`mindspore.device_context.gpu.op_precision.conv_allow_tf32`	Whether to convert FP32 to TF32 for Conv operators.
`mindspore.device_context.gpu.op_precision.matmul_allow_tf32`	Whether to convert FP32 to TF32 for Matmul operators.
`mindspore.device_context.gpu.op_tuning.conv_dgrad_algo`	Specifies convolution data grad algorithm.
`mindspore.device_context.gpu.op_tuning.conv_fprop_algo`	Specifies convolution forward algorithm.
`mindspore.device_context.gpu.op_tuning.conv_wgrad_algo`	Specifies convolution filter grad algorithm.

`mindspore.device_context.ascend.device_count`	Returns the compute-capable device count of Ascend.
`mindspore.device_context.ascend.is_available`	Returns whether the Ascend backend is available.
`mindspore.device_context.ascend.op_precision.conv_allow_hf32`	Whether to convert FP32 to HF32 for Conv operators.
`mindspore.device_context.ascend.op_precision.matmul_allow_hf32`	Whether to convert FP32 to HF32 for Matmul operators.
`mindspore.device_context.ascend.op_precision.precision_mode`	Configure mixed precision mode setting.
`mindspore.device_context.ascend.op_precision.op_precision_mode`	Path to the configuration file for op precision mode.
`mindspore.device_context.ascend.op_debug.execute_timeout`	Set the maximum duration of executing an operator in seconds.
`mindspore.device_context.ascend.op_debug.debug_option`	Enable debugging options for Ascend operators, disabled by default.
`mindspore.device_context.ascend.op_debug.aclinit_config`	Configure the configuration items for the aclInit interface.
`mindspore.device_context.ascend.op_tuning.op_compile`	Whether to select online compilation.
`mindspore.device_context.ascend.op_tuning.aclnn_cache`	Configure aclnn cache parameters.
`mindspore.device_context.ascend.op_tuning.aoe_tune_mode`	AOE tuning mode setting, which is not set by default.
`mindspore.device_context.ascend.op_tuning.aoe_job_type`	Set the parameters specific to Ascend Optimization Engine.