# tahahah/sana-fork architecture

> Architecture diagram of the GitHub repository tahahah/sana-fork, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2025-01-29). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/tahahah/sana-fork
- Repository: https://github.com/tahahah/sana-fork

## Overview

##### Main Components (Top Level)

**A. Input Layer**
- Text Input (Prompts)
- Configuration Settings

**B. Core Processing Layer (Main Architecture)**
- DC-AE (Autoencoder)
- Linear DiT (Diffusion Transformer)
- Decoder-only Text Encoder
- Flow-DPM-Solver

**C. Output Layer**
- Generated Images (up to 4096×4096)

##### Detailed Component Breakdown

a) Data Flow Components:
- Data Pipeline (/diffusion/data/)
- → Dataset Handlers
- → Transforms
- → WIDS (Data Loading System)

- Model Components (/diffusion/model/)
- → DC-AE Module
- → Linear Attention Modules
- → Diffusion Utils
- → Gaussian Diffusion

b) Training Pipeline:
- Training Flow
- → Optimizer
- → Scheduler
- → Checkpoint Management
- → Distributed Training Support

c) Inference Pipeline:
- Inference System
- → Sampling Mechanisms
- → Pipeline Orchestration
- → Multi-threading Support

##### Integration Points

- Data flow between components
- Training/Inference paths
- Feedback loops
- Optimization paths

##### Technical Details

- Model Sizes (0.6B and 1.6B variants)
- Resolution Support (512×512, 1024×1024, 4096×4096)
- Performance Metrics
- Hardware Requirements

##### Key Implementation Notes

Annotations for:
- Linear Attention Optimization
- 32× Image Compression
- Flow-DPM-Solver Integration
- Multi-GPU Support

##### External Integrations

Connections to:
- Metrics Tools (/tools/metrics/)
- Evaluation Systems
- Model Export Utilities

- Docker Container Support
- API Endpoints
- Resource Requirements
- Scaling Capabilities

- FID scores
- CLIP scores
- GenEval metrics
- DPG bench results

## Diagram (Mermaid)

```mermaid
flowchart TB
    subgraph Data["Data Pipeline"]
        direction TB
        ImgText[("Image-Text Pairs")]
        Config["Config Files"]:::config
        Dataset["SanaImgDataset"]:::dataset
        Transform["Data Transforms"]:::transform
        Builder["Data Builder"]:::builder
        
        ImgText --> Dataset
        Config --> Builder
        Dataset --> Transform
        Transform --> Builder
    end

    subgraph Models["Model Components"]
        direction TB
        DCAE["DC-AE Compression"]:::dcae
        DiT["Linear DiT"]:::dit
        TextEnc["Text Encoder"]:::model
        DPM["Flow-DPM-Solver"]:::dpm
    end

    subgraph Training["Training Pipeline"]
        direction TB
        Forward["Forward Pass"]
        
        subgraph Loss["Loss Calculation"]
            direction TB
            DiffLoss["Diffusion Loss"]:::loss
            AlignLoss["Text-Image Alignment Loss"]:::loss
            CombLoss["Combined Loss"]:::loss
            
            DiffLoss --> CombLoss
            AlignLoss --> CombLoss
        end
        
        subgraph Opt["Optimization"]
            direction TB
            Optimizer["AdamW Optimizer"]:::opt
            LRSched["Learning Rate Scheduler"]:::opt
            Update["Model Updates"]
        end
        
        Checkpoint["Checkpointing"]:::ckpt
    end

    %% Main Flow
    Builder --> DCAE
    DCAE --> DiT
    Builder --> TextEnc
    TextEnc --> DiT
    DiT --> Forward
    Forward --> DiffLoss
    Forward --> AlignLoss
    CombLoss --> Optimizer
    Optimizer --> Update
    LRSched --> Optimizer
    Update --> Checkpoint

    %% Styles
    classDef config fill:#e6f3ff,stroke:#4a90e2
    classDef dataset fill:#e6f3ff,stroke:#4a90e2
    classDef transform fill:#e6f3ff,stroke:#4a90e2
    classDef builder fill:#e6f3ff,stroke:#4a90e2
    classDef model fill:#d4edda,stroke:#28a745
    classDef dcae fill:#d4edda,stroke:#28a745
    classDef dit fill:#d4edda,stroke:#28a745
    classDef dpm fill:#d4edda,stroke:#28a745
    classDef loss fill:#f8d7da,stroke:#dc3545
    classDef opt fill:#fff3cd,stroke:#ffc107
    classDef ckpt fill:#e2e3e5,stroke:#6c757d

    %% Click Events
    click Config "https://github.com/tahahah/sana-fork/tree/main/configs/sana_config/"
    click Dataset "https://github.com/tahahah/sana-fork/blob/main/diffusion/data/datasets/sana_data.py"
    click Transform "https://github.com/tahahah/sana-fork/blob/main/diffusion/data/transforms.py"
    click Builder "https://github.com/tahahah/sana-fork/blob/main/diffusion/data/builder.py"
    click DCAE "https://github.com/tahahah/sana-fork/tree/main/diffusion/model/dc_ae/"
    click DiT "https://github.com/tahahah/sana-fork/blob/main/diffusion/model/nets/sana.py"
    click DPM "https://github.com/tahahah/sana-fork/blob/main/diffusion/dpm_solver.py"
    click Forward "https://github.com/tahahah/sana-fork/blob/main/train_scripts/train.py"
    click Optimizer "https://github.com/tahahah/sana-fork/blob/main/diffusion/utils/optimizer.py"
    click LRSched "https://github.com/tahahah/sana-fork/blob/main/diffusion/utils/lr_scheduler.py"
    click Checkpoint "https://github.com/tahahah/sana-fork/blob/main/diffusion/utils/checkpoint.py"
```
