integrations.diffusion.plugin

integrations.diffusion.plugin

Diffusion LM training plugin for Axolotl.

Classes

Name Description
DiffusionPlugin Plugin for diffusion language model training.

DiffusionPlugin

integrations.diffusion.plugin.DiffusionPlugin()

Plugin for diffusion language model training.

This plugin enables diffusion-based training using the LLaDA approach, which uses random masking and bidirectional attention to train language models.

Methods

Name Description
get_input_args Returns the pydantic model for LLaDA plugin arguments.
get_trainer_cls Return custom trainer class for diffusion training.
post_model_load Perform actions after model is loaded.
post_trainer_create Configure trainer after creation.
get_input_args
integrations.diffusion.plugin.DiffusionPlugin.get_input_args()

Returns the pydantic model for LLaDA plugin arguments.

get_trainer_cls
integrations.diffusion.plugin.DiffusionPlugin.get_trainer_cls(cfg)

Return custom trainer class for diffusion training.

post_model_load
integrations.diffusion.plugin.DiffusionPlugin.post_model_load(cfg, model)

Perform actions after model is loaded.

post_trainer_create
integrations.diffusion.plugin.DiffusionPlugin.post_trainer_create(cfg, trainer)

Configure trainer after creation.