bridge.diffusion.models.flux

FLUX diffusion model implementation for DFM.

This module provides the FLUX model architecture, which is a state-of-the-art text-to-image diffusion model using MMDiT-style transformer blocks.

Components: - Flux: Main FLUX model class - FluxProvider: Configuration and provider dataclass for FLUX models - MMDiTLayer: Multi-modal DiT layer for double blocks - FluxSingleTransformerBlock: Single transformer block for FLUX - JointSelfAttention: Joint self-attention for MMDiT layers - FluxSingleAttention: Self-attention for single blocks - EmbedND: N-dimensional rotary position embedding - MLPEmbedder: MLP embedding module - TimeStepEmbedder: Timestep embedding module - AdaLN: Adaptive Layer Normalization - AdaLNContinuous: Continuous Adaptive Layer Normalization

Subpackages

• bridge.diffusion.models.flux.flow_matching
  • bridge.diffusion.models.flux.flow_matching.flux_inference_pipeline
  • bridge.diffusion.models.flux.flow_matching.flux_adapter

Submodules

• bridge.diffusion.models.flux.flux_attention
• bridge.diffusion.models.flux.flux_provider
• bridge.diffusion.models.flux.flux_layer_spec
• bridge.diffusion.models.flux.layers
• bridge.diffusion.models.flux.flux_step
• bridge.diffusion.models.flux.flux_model

Package Contents

Data

__all__

API

bridge.diffusion.models.flux.__all__

[‘Flux’, ‘FluxProvider’, ‘MMDiTLayer’, ‘FluxSingleTransformerBlock’, ‘JointSelfAttention’, ‘JointSel…