spikingjelly.clock_driven.monitor package

Module contents

class spikingjelly.clock_driven.monitor.Monitor(net: Module, device: Optional[str] = None, backend: str = 'numpy')[源代码]

基类：object

参数

  • net (nn.Module) – 要监视的网络

  • device (str, optional) – 监视数据的存储和处理的设备，仅当backend为 'torch' 时有效。可以为 'cpu', 'cuda', 'cuda:0' 字符串或者 torch.device 类型，默认为 None

  • backend (str, optional) – 监视数据的处理后端。可以为 'torch', 'numpy' ，默认为 'numpy'

参数

  • net (nn.Module) – Network to be monitored

  • device (str, optional) – Device carrying and processing monitored data. Only take effect when backend is set to 'torch'. Can be string 'cpu', 'cuda', 'cuda:0' or torch.device, defaults to None

  • backend (str, optional) – Backend processing monitored data, can be 'torch', 'numpy', defaults to 'numpy'

enable()[源代码]

启用Monitor的监视功能，开始记录数据

Enable Monitor. Start recording data.

disable()[源代码]

禁用Monitor的监视功能，不再记录数据

Disable Monitor. Stop recording data.

forward_hook(module, input, output)[源代码]
reset()[源代码]

清空之前的记录数据

Delete previously recorded data

get_avg_firing_rate(all: bool = True, module_name: Optional[str] = None) → Tensor[源代码]
参数

  • all (bool, optional) – 是否为所有层的总平均发放率，默认为 True

  • module_name (str, optional) – 层的名称，仅当all为 False 时有效

返回

所关心层的平均发放率

返回类型

torch.Tensor or float

参数

  • all (bool, optional) – Whether needing firing rate averaged on all layers, defaults to True

  • module_name (str, optional) – Name of concerned layer. Only take effect when all is False

返回

Averaged firing rate on concerned layers

返回类型

torch.Tensor or float

get_nonfire_ratio(all: bool = True, module_name: Optional[str] = None) → Tensor[源代码]
参数

  • all (bool, optional) – 是否为所有层的静默神经元比例，默认为 True

  • module_name (str, optional) – 层的名称，仅当all为 False 时有效

返回

所关心层的静默神经元比例

返回类型

torch.Tensor or float

参数

  • all (bool, optional) – Whether needing ratio of silent neurons of all layers, defaults to True

  • module_name (str, optional) – Name of concerned layer. Only take effect when all is False

返回

Ratio of silent neurons on concerned layers

返回类型

torch.Tensor or float