import com.mindspore.config.MSContext;
MSContext is defined for holding environment variables during runtime.
public boolean init()
Use default parameters initialize MSContext, use two thread, no bind, no parallel.
Returns
Whether the initialization is successful.
public boolean init(int threadNum, int cpuBindMode)
Initialize MSContext for cpu.
Parameters
threadNum
: Thread number config for thread pool.cpuBindMode
: A CpuBindMode enum variable.Returns
Whether the initialization is successful.
public boolean init(int threadNum, int cpuBindMode, boolean isEnableParallel)
Initialize MSContext.
Parameters
threadNum
: Thread number config for thread pool.cpuBindMode
: A CpuBindMode enum variable.isEnableParallel
: Is enable parallel in different device.Returns
Whether the initialization is successful.
public boolean addDeviceInfo(int deviceType, boolean isEnableFloat16)
Add device info for mscontext.
Parameters
deviceType
: A DeviceType enum type.isEnableFloat16
: Is enable fp16.Returns
Whether the device info add is successful.
boolean addDeviceInfo(int deviceType, boolean isEnableFloat16, int npuFreq)
Add device info for mscontext.
Parameters
deviceType
: A DeviceType enum type.isEnableFloat16
: is enable fp16.npuFreq
: Npu frequency.Returns
Whether the device info add is successful.
public long getMSContextPtr()
Get MSContext pointer.
Returns
The MSContext pointer.
public void free()
Free the memory allocated for MSContext.
void setThreadNum(int threadNum)
Sets the number of threads at runtime. Only valid for Lite. If MSContext is not initialized, this function will do nothing and output null pointer information in the log.
Parameters
threadNum
: Number of threads at runtime.void int getThreadNum()
Get the current thread number setting. Only valid for Lite. If MSContext is not initialized, this function will do nothing and output null pointer information in the log.
Returns
The current thread number setting.
void setInterOpParallelNum(int parallelNum)
Set the parallel number of operators at runtime. Only valid for Lite. If MSContext is not initialized, this function will do nothing and output null pointer information in the log.
Parameters
parallelNum
: the parallel number of operators at runtime.int getInterOpParallelNum()
et the current operators parallel number setting. Only valid for Lite. If MSContext is not initialized, this function will return -1 and output null pointer information in the log.
Returns
The current operators parallel number setting.
void setThreadAffinity(int mode)
Set the thread affinity to CPU cores. Only valid for Lite. If MSContext is not initialized, this function will do nothing and output null pointer information in the log.
Parameters
mode
: 0, no affinities; 1, big cores first; 2, little cores first. int getThreadAffinityMode()
Get the thread affinity of CPU cores. Only valid for Lite. If MSContext is not initialized, this function will return -1 and output null pointer information in the log.
Returns
Thread affinity to CPU cores. 0, no affinities; 1, big cores first; 2, little cores first.
void setThreadAffinity(ArrayList<Integer> coreList)
Set the thread lists to CPU cores, if two different setThreadAffinity
are set for a single MSContext at the same time,only coreList
will take effect and mode
will not. Only valid for Lite.
If MSContext is not initialized, this function will do nothing and output null pointer information in the log.
Parameters
coreList
: An Arraylist of thread core list.ArrayList<Integer> getThreadAffinityCoreList()
Get the thread lists of CPU cores. Only valid for Lite. If MSContext is not initialized, this function will retutn an empty Arraylist and output null pointer information in the log.
Returns
An Arraylist of thread core list.
void setEnableParallel(boolean isParallel)
Set the status whether to perform model inference or training in parallel. Only valid for Lite. If MSContext is not initialized, this function will do nothing and output null pointer information in the log.
Parameters
isParallel
: true, parallel; false, not in parallel.boolean getEnableParallel()
Get the status whether to perform model inference or training in parallel. Only valid for Lite.
If MSContext is not initialized, this function will return false
and output null pointer information in the log.
Returns
Bool value that indicates whether in parallel. true, parallel; false, not in parallel.
import com.mindspore.config.DeviceType;
Define device type.
public static final int DT_CPU = 0;
public static final int DT_GPU = 1;
public static final int DT_NPU = 2;
The value of DeviceType is 0, and the specified device type is CPU.
The value of DeviceType is 1, and the specified device type is GPU.
The value of DeviceType is 2, and the specified device type is NPU.
import com.mindspore.config.CpuBindMode;
Define CPU core bind mode.
public static final int MID_CPU = 2;
public static final int HIGHER_CPU = 1;
public static final int NO_BIND = 0;
The value of CpuBindMode is 2, and the middle core is bound first.
The value of CpuBindMode is 1, and higher cores are bound first.
The value of CpuBindMode is 0, and no core is bound.
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。