fiftyone.utils.torch.SaveLayerTensor

class documentation

class SaveLayerTensor(object): (source)

Callback that saves the input/output tensor of the specified layer of a Torch model during each forward() call.

Parameters
model	the Torch model, a `torch:torch.nn.Module`
layer_name	the name of the layer whose output to save. Prepend `"<"` to save the input tensor instead

Method	`__call__`	Undocumented
Method	`__init__`	Undocumented
Property	`tensor`	The tensor saved from the last `forward()` call.
Method	`_setup`	Undocumented
Instance Variable	`_save_input`	Undocumented
Instance Variable	`_tensor`	Undocumented

def __call__(self, module, module_in, module_out): (source) ¶

Undocumented

def __init__(self, model, layer_name): (source) ¶

Undocumented

The tensor saved from the last forward() call.

def _setup(self, model, layer_name): (source) ¶

Undocumented

Undocumented

Undocumented