# rosettacommons/rfdiffusion architecture

> Architecture diagram of the GitHub repository rosettacommons/rfdiffusion, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-03). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/rosettacommons/rfdiffusion
- Repository: https://github.com/rosettacommons/rfdiffusion

## Overview

RFdiffusion generates protein structures through configurable diffusion inference: a user launches the inference script, which prepares a sampler and conditions, runs repeated denoising with the learned structure model, and writes designs. The README documents unconditional and motif-conditioned generation, symmetry, binder design, fold conditioning, and partial diffusion; these are represented as sampler capabilities, though their detailed implementations are not all shown. The graph also shows the RoseTTAFold model layers and an internal SE(3)-Transformer layer. Links are limited to README statements and source-index references; unsampled implementation details are not asserted.

Main entry points: [`rfdiffusion/inference/model_runners.py`](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/model_runners.py) (Sampling runners), [`rfdiffusion/Track_module.py`](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Track_module.py) (Track module), [`rfdiffusion/RoseTTAFoldModel.py`](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/RoseTTAFoldModel.py) (RoseTTAFold model), [`rfdiffusion/Attention_module.py`](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Attention_module.py) (Attention layers)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_inference["Inference orchestration"]
  node_entry["Inference script<br/>[run_inference.py]"]
  node_inference_utils["Inference utilities<br/>[utils.py]"]
  node_sampler["Sampling runners<br/>[model_runners.py]"]
end

subgraph group_conditioning["Design conditioning"]
  node_contigs["Contig mapping<br/>[contigs.py]"]
  node_symmetry["Symmetry support<br/>[symmetry.py]"]
  node_pot_manager["Potential manager<br/>[manager.py]"]
  node_potentials["Guidance potentials<br/>[potentials.py]"]
end

subgraph group_generation["Structure generation"]
  node_diffuser["Diffusion process<br/>[diffusion.py]"]
  node_so3["SO&#40;3&#41; schedules<br/>[igso3.py]"]
  node_geometry["Protein geometry<br/>[kinematics.py]"]
  node_chemistry["Residue chemistry<br/>[chemical.py]"]
  node_output["Design outputs<br/>[util.py]"]
end

subgraph group_network["Learned model"]
  node_rose["RoseTTAFold model"]
  node_embeddings["Feature embeddings<br/>[Embeddings.py]"]
  node_track["Track module<br/>[Track_module.py]"]
  node_attention["Attention layers"]
  node_se3["SE&#40;3&#41; network<br/>[SE3_network.py]"]
  node_tensor_ops["Tensor utilities<br/>[util_module.py]"]
  node_se3_transformer["SE&#40;3&#41; Transformer<br/>[transformer.py]"]
  node_fiber["Fiber features<br/>[fiber.py]"]
  node_se3_convolution["Equivariant convolution<br/>[convolution.py]"]
end

node_user(("Design user"))

node_user -->|"launches"| node_entry
node_entry -->|"uses"| node_inference_utils
node_inference_utils -->|"references"| node_sampler
node_sampler -->|"uses"| node_contigs
node_sampler -->|"uses"| node_symmetry
node_sampler -->|"configures"| node_pot_manager
node_pot_manager -->|"creates"| node_potentials
node_sampler -->|"uses"| node_diffuser
node_diffuser -->|"uses"| node_so3
node_sampler -->|"uses"| node_geometry
node_sampler -->|"uses"| node_chemistry
node_sampler -->|"runs model"| node_rose
node_sampler -.->|"writes designs"| node_output
node_rose -->|"uses"| node_embeddings
node_rose -->|"uses"| node_track
node_embeddings -->|"uses"| node_attention
node_track -->|"uses"| node_attention
node_track -->|"uses"| node_se3
node_attention -->|"uses"| node_tensor_ops
node_track -->|"uses"| node_tensor_ops
node_se3_transformer -->|"uses"| node_fiber
node_se3_transformer -->|"uses"| node_se3_convolution

click node_entry "https://github.com/rosettacommons/rfdiffusion/blob/main/scripts/run_inference.py"
click node_inference_utils "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/utils.py"
click node_sampler "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/model_runners.py"
click node_contigs "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/contigs.py"
click node_symmetry "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/symmetry.py"
click node_pot_manager "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/potentials/manager.py"
click node_potentials "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/potentials/potentials.py"
click node_diffuser "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/diffusion.py"
click node_so3 "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/igso3.py"
click node_geometry "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/kinematics.py"
click node_chemistry "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/chemical.py"
click node_output "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/util.py"
click node_rose "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/RoseTTAFoldModel.py"
click node_embeddings "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Embeddings.py"
click node_track "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Track_module.py"
click node_attention "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Attention_module.py"
click node_se3 "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/SE3_network.py"
click node_tensor_ops "https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/util_module.py"
click node_se3_transformer "https://github.com/rosettacommons/rfdiffusion/blob/main/env/SE3Transformer/se3_transformer/model/transformer.py"
click node_fiber "https://github.com/rosettacommons/rfdiffusion/blob/main/env/SE3Transformer/se3_transformer/model/fiber.py"
click node_se3_convolution "https://github.com/rosettacommons/rfdiffusion/blob/main/env/SE3Transformer/se3_transformer/model/layers/convolution.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_entry,node_inference_utils,node_sampler,node_user toneBlue
class node_contigs,node_symmetry,node_pot_manager,node_potentials toneAmber
class node_diffuser,node_so3,node_geometry,node_chemistry,node_output toneMint
class node_rose,node_embeddings,node_track,node_attention,node_se3,node_tensor_ops,node_se3_transformer,node_fiber,node_se3_convolution toneRose
```

## Components

### Inference orchestration

- [Inference script](https://github.com/rosettacommons/rfdiffusion/blob/main/scripts/run_inference.py): `scripts/run_inference.py`
- [Inference utilities](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/utils.py): `rfdiffusion/inference/utils.py`
- [Sampling runners](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/model_runners.py): `rfdiffusion/inference/model_runners.py`

### Design conditioning

- [Contig mapping](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/contigs.py): `rfdiffusion/contigs.py`
- [Symmetry support](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/inference/symmetry.py): `rfdiffusion/inference/symmetry.py`
- [Potential manager](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/potentials/manager.py): `rfdiffusion/potentials/manager.py`
- [Guidance potentials](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/potentials/potentials.py): `rfdiffusion/potentials/potentials.py`

### Structure generation

- [Diffusion process](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/diffusion.py): `rfdiffusion/diffusion.py`
- [SO(3) schedules](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/igso3.py): `rfdiffusion/igso3.py`
- [Protein geometry](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/kinematics.py): `rfdiffusion/kinematics.py`
- [Residue chemistry](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/chemical.py): `rfdiffusion/chemical.py`
- [Design outputs](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/util.py): `rfdiffusion/util.py`

### Learned model

- [RoseTTAFold model](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/RoseTTAFoldModel.py): `rfdiffusion/RoseTTAFoldModel.py`
- [Feature embeddings](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Embeddings.py): `rfdiffusion/Embeddings.py`
- [Track module](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Track_module.py): `rfdiffusion/Track_module.py`
- [Attention layers](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/Attention_module.py): `rfdiffusion/Attention_module.py`
- [SE(3) network](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/SE3_network.py): `rfdiffusion/SE3_network.py`
- [Tensor utilities](https://github.com/rosettacommons/rfdiffusion/blob/main/rfdiffusion/util_module.py): `rfdiffusion/util_module.py`
- [SE(3) Transformer](https://github.com/rosettacommons/rfdiffusion/blob/main/env/SE3Transformer/se3_transformer/model/transformer.py): `env/SE3Transformer/se3_transformer/model/transformer.py`
- [Fiber features](https://github.com/rosettacommons/rfdiffusion/blob/main/env/SE3Transformer/se3_transformer/model/fiber.py): `env/SE3Transformer/se3_transformer/model/fiber.py`
- [Equivariant convolution](https://github.com/rosettacommons/rfdiffusion/blob/main/env/SE3Transformer/se3_transformer/model/layers/convolution.py): `env/SE3Transformer/se3_transformer/model/layers/convolution.py`

### Other components

- **Design user**

## Connections

- Design user → Inference script: launches (evidence: `README.md`)
- Inference script → Inference utilities: uses (evidence: `scripts/run_inference.py`)
- Inference utilities → Sampling runners: references (evidence: `rfdiffusion/inference/utils.py`)
- Sampling runners → Contig mapping: uses (evidence: `rfdiffusion/inference/model_runners.py`)
- Sampling runners → Symmetry support: uses (evidence: `rfdiffusion/inference/model_runners.py`)
- Sampling runners → Potential manager: configures (evidence: `rfdiffusion/inference/model_runners.py`)
- Potential manager → Guidance potentials: creates (evidence: `rfdiffusion/potentials/manager.py`)
- Sampling runners → Diffusion process: uses (evidence: `rfdiffusion/inference/model_runners.py`)
- Diffusion process → SO(3) schedules: uses (evidence: `rfdiffusion/diffusion.py`)
- Sampling runners → Protein geometry: uses (evidence: `rfdiffusion/inference/model_runners.py`)
- Sampling runners → Residue chemistry: uses (evidence: `rfdiffusion/inference/model_runners.py`)
- Sampling runners → RoseTTAFold model: runs model (evidence: `rfdiffusion/inference/model_runners.py`)
- Sampling runners → Design outputs: writes designs (evidence: `rfdiffusion/inference/model_runners.py`)
- RoseTTAFold model → Feature embeddings: uses (evidence: `rfdiffusion/RoseTTAFoldModel.py`)
- RoseTTAFold model → Track module: uses (evidence: `rfdiffusion/RoseTTAFoldModel.py`)
- Feature embeddings → Attention layers: uses (evidence: `rfdiffusion/Embeddings.py`)
- Track module → Attention layers: uses (evidence: `rfdiffusion/Track_module.py`)
- Track module → SE(3) network: uses (evidence: `rfdiffusion/Track_module.py`)
- Attention layers → Tensor utilities: uses (evidence: `rfdiffusion/Attention_module.py`)
- Track module → Tensor utilities: uses (evidence: `rfdiffusion/Track_module.py`)
- SE(3) Transformer → Fiber features: uses (evidence: `env/SE3Transformer/se3_transformer/model/transformer.py`)
- SE(3) Transformer → Equivariant convolution: uses (evidence: `env/SE3Transformer/se3_transformer/model/transformer.py`)
