bridge.diffusion.recipes.flux.flux

Module Contents

Functions

flux_12b_pretrain_config	Return a pre-training configuration for FLUX 12B model.
flux_12b_sft_config	Return an SFT (supervised fine-tuning) configuration for FLUX 12B model.

API

bridge.diffusion.recipes.flux.flux.flux_12b_pretrain_config() → megatron.bridge.training.config.ConfigContainer

Return a pre-training configuration for FLUX 12B model.

Default parallelism: TP=2, PP=1. Uses mock/synthetic data when data_paths is None. To customize (e.g. data paths, checkpoint dir), edit this recipe or add a new recipe that builds on these defaults.

bridge.diffusion.recipes.flux.flux.flux_12b_sft_config(

pretrained_checkpoint: str | None = None,

) → megatron.bridge.training.config.ConfigContainer

Return an SFT (supervised fine-tuning) configuration for FLUX 12B model.

Uses the same defaults as flux_12b_pretrain_config() and overrides checkpoint to load from pretrained_checkpoint when provided.