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 module
max.pipelines.lib.role_contracts
The inputs and outputs of each role a registered module fills.
An architecture registers a module class per role in
modules: a
Module (ModuleV3) or a graph-API
Module. The framework builds the role’s graph and
batches from the contract here, so the module’s forward (__call__ for
a graph-API module) takes the role’s inputs and returns its outputs. A
ModuleV3 module sees Tensor fields, and a graph-API
module TensorValue fields.
Decoder
DecoderInputs | The inputs of a DECODER module's forward. |
|---|---|
DecoderOutputs | The outputs of a DECODER module's forward. |
Vision
VisionInputs | The image embeddings a DECODER merges into its token stream. |
|---|