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

Python class

HiddenStates

HiddenStates

class max.nn.graph_groups.HiddenStates(*args, **kwargs)

source

Bases: GraphGroup[str, tuple[TensorValue, …]]

Hidden states a decoder returns for a drafter, a capture, or the next pipeline stage, keyed by the one mode it ran in.

Parameters:

• args (object)
• kwargs (object)

Return type:

Self

kind

kind: ClassVar[str | None] = 'hidden_states'

source

Identifies the feature wherever the group sits in a tree.

tensors()

tensors()

source

Returns the hidden states, whichever mode produced them.

Return type:

tuple[TensorValue, …]