For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Python class

PrintHook

`PrintHook`

class max.nn.hooks.PrintHook(export_path=None, filter=None)

source

Bases: BasePrintHook

Hook that prints layer tensor inputs and outputs.

Initialize the class before building the graph. This way, the hook can add print ops to the graph.

Parameters:

export_path (str | None) – Intended as the directory to write recorded tensors into, but writing to disk isn’t yet supported on the MAX execution path. When set, the path is ignored.
filter (list[str] | None) – A list of layer names to restrict printing to. When None, every named layer is printed.

`export_path`

property export_path: str | None

source

`name_layers()`

name_layers(model)

source

Create names for all layers in the model based on nested attributes.

Parameters:: model (Layer | Module[..., Any]) – A V2 Layer or V3 Module instance.
Raises:: TypeError – If model is neither a V2 Layer nor a V3 Module.
Return type:: None

`name_layers_v3()`

name_layers_v3(model)

source

Name all v3 Module layers and wrap their forward to fire the hook.

Parameters:: model (Module[..., Any])
Return type:: None

`print_value()`

print_value(name, value)

source

Prints a value, and returns whether the print is successful.

Parameters:

name (str)
value (Any)

Return type:

bool

`remove()`

remove()

source

Return type:: None

PrintHook​

export_path​

name_layers()​

name_layers_v3()​

print_value()​

remove()​