mindspore.profiler.schedule
- class mindspore.profiler.schedule(*, wait: int, active: int, warmup: int = 0, repeat: int = 0, skip_first: int = 0)[源代码]
该类用于获取每一步的操作。
调度如下:
(NONE) (NONE) (NONE) (WARM_UP) (RECORD) (RECORD) (RECORD_AND_SAVE) None START------->skip_first------->wait-------->warmup-------->active........active.........active----------->stop | | | repeat_1 | ---------------------------------------------------------------
Profiler将跳过前
skip_first
步,然后等待wait
步,接着在接下来的warmup
步中进行预热,然后在接下来的active
步中进行活动记录,然后从wait
步开始重复循环。可选的循环次数由repeat
参数指定,repeat
值为0表示循环将继续直到分析完成。- 关键字参数:
wait (int) - 预热阶段等待的步数,必须大于等于0。如果外部不设置wait参数,会在初始化schedule类时,设置为
0
。active (int) - 活动阶段执行的步数,必须大于等于1。如果外部不设置active参数,会在初始化schedule类时,设置为
1
。warmup (int, 可选) - 预热阶段执行的步数,必须大于等于0。默认值:
0
。repeat (int, 可选) - 重复次数,必须大于等于0。如果repeat设置为0,Profiler会根据模型训练次数来确定repeat值,例如总训练步数为100,wait+active+warmup=10,skip_first=10,则repeat=(100-10)/10=9,表示重复执行9次,但此时会多生成一个采集不完整的的性能数据,最后一个step的数据用户无需关注,为异常数据。建议配置大于0的整数。当使用集群分析工具或MindStudio Insight查看时,建议配置为1;若设置大于1,则需要将采集的性能数据文件夹分为repeat等份,放到不同文件夹下重新解析,分类方式按照文件夹名称中的时间戳先后。默认值:
0
。skip_first (int, 可选) - 跳过第一个步数。默认值:
0
。
- 异常:
ValueError - 参数step必须大于等于0。
- 支持平台:
Ascend
样例:
>>> import numpy as np >>> import mindspore >>> import mindspore.dataset as ds >>> from mindspore import context, nn >>> from mindspore.profiler import ProfilerLevel, AicoreMetrics, ExportType, ProfilerActivity >>> >>> class Net(nn.Cell): ... def __init__(self): ... super(Net, self).__init__() ... self.fc = nn.Dense(2, 2) ... ... def construct(self, x): ... return self.fc(x) >>> >>> def generator_net(): ... for _ in range(2): ... yield np.ones([2, 2]).astype(np.float32), np.ones([2]).astype(np.int32) >>> >>> def train(test_net): ... optimizer = nn.Momentum(test_net.trainable_params(), 1, 0.9) ... loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True) ... data = ds.GeneratorDataset(generator_net(), ["data", "label"]) ... model = mindspore.train.Model(test_net, loss, optimizer) ... model.train(1, data) >>> >>> if __name__ == '__main__': ... # If the device_target is GPU, set the device_target to "GPU" ... context.set_context(mode=mindspore.GRAPH_MODE) ... mindspore.set_device("Ascend") ... ... # Init Profiler ... experimental_config = mindspore.profiler._ExperimentalConfig( ... profiler_level=ProfilerLevel.Level0, ... aic_metrics=AicoreMetrics.AiCoreNone, ... l2_cache=False, ... mstx=False, ... data_simplification=False, ... export_type=[ExportType.Text]) ... steps = 10 ... net = Net() ... # Note that the Profiler should be initialized before model.train ... with mindspore.profiler.profile(activities=[ProfilerActivity.CPU, ProfilerActivity.NPU], ... schedule=mindspore.profiler.schedule(wait=1, warmup=1, active=2, ... repeat=1, skip_first=2), ... on_trace_ready=mindspore.profiler.tensorboard_trace_handler("./data"), ... profile_memory=False, ... experimental_config=experimental_config) as prof: ... ... # Train Model ... for step in range(steps): ... train(net) ... prof.step()
- to_dict()
将schedule类转换为一个字典。
- 返回:
字典,schedule类的参数与对应的值。