Spaces:
Running on Zero
Running on Zero
| """Flux2 transformer model components for LightDiffusion-Next. | |
| This module contains the Flux2 model architecture including: | |
| - DoubleStreamBlock: Joint image-text transformer blocks | |
| - SingleStreamBlock: Merged stream transformer blocks | |
| - Flux2: Main model class with forward pass | |
| Ported from ComfyUI's Flux2 Klein implementation. | |
| """ | |
| from src.NeuralNetwork.flux2.model import Flux2, Flux2Params | |
| from src.NeuralNetwork.flux2.layers import ( | |
| DoubleStreamBlock, | |
| SingleStreamBlock, | |
| LastLayer, | |
| MLPEmbedder, | |
| EmbedND, | |
| QKNorm, | |
| ) | |
| __all__ = [ | |
| "Flux2", | |
| "Flux2Params", | |
| "DoubleStreamBlock", | |
| "SingleStreamBlock", | |
| "LastLayer", | |
| "MLPEmbedder", | |
| "EmbedND", | |
| "QKNorm", | |
| ] | |