Skip to content
Adam-AtlasSoftwarePublic

About

Radeon Decoupled Neural Upscaler - RDNA3-focused ML-based uUpscaler using sunny2109/RDG (Decoupled G-buffer Guidance). Pretrained on guluthemonster/M3VIR and LLLebin/GameIR, fine-tuned on vKITTI2 and MPI-Sintel.

Resources

Stars

16 stars

Watchers

6 watching

Forks

Repository files navigation

Radeon Decoupled Neural Upscaler (RDNU)

An open ML temporal upscaler for AMD RDNA2 and RDNA3 GPUs, in the class of FSR 4 and DLSS: jittered low-resolution colour, depth and motion vectors in, anti-aliased high-resolution frames out. The runtime is DirectX 12 compute with INT8 networks (DP4a on RDNA2, wave-matrix on RDNA3) and exposes the FidelityFX API so it can drop into the FSR sample or be injected into games.

The network is Arm Neural Super Sampling (148K params, ~2.8K MAC per output pixel at 2×), chosen after the review in docs/Model_Candidates.md. The earlier RDG-based model (CNN with G-buffer guidance, trained here on GameIR/M3VIR/vKITTI2/Sintel) is kept under RDG/, scripts/ and runtime/models/ as the reference baseline.

Status

  • NSS backbone exported to the engine's bundle format with per-layer goldens (fp32 and INT8 QAT); the harness graph for it is written and awaits GPU validation.
  • The RDG network runs end to end on the GPU in the harness (bit-exact INT8, WMMA 1×1 conv).
  • Not yet: the NSS pre/post passes in HLSL, the DX12 backend for Arm's NSS component, sample integration. Order of work: docs/NSS_Runtime_Plan.md.

Layout

RDNU/
├── docs/                 # plans and reviews (start with NSS_Runtime_Plan.md)
├── ml/
│   ├── model-gym/        # Arm Neural Graphics Model Gym (train / fine-tune / QAT / export)   [submodule]
│   ├── capture-unreal/   # UE 5.5 dataset capture plugin                                     [submodule]
│   ├── models/nss, nfru  # released weights (Git LFS)                                        [submodules]
│   └── data/             # datasets, opt-in submodules (see docs/Datasets.md)
├── runtime/
│   ├── harness/          # standalone DX12 engine + kernels, validated layer by layer vs golden
│   ├── tools/            # nss_export.py (NSS bundles), dump_golden.py (RDG bundles)
│   ├── src/, shaders/    # FSR-sample backend and its placeholder shaders (to be replaced)
│   └── external/         # vcpkg, DirectX-Headers, FidelityFX SDK fork, Arm Neural Graphics SDK
├── RDG/, scripts/        # RDG baseline: training code, configs, utilities
└── CMakeLists.txt, CMakePresets.json

Setup

git clone --recurse-submodules git@github.com:Adam-AtlasSoftware/RDNU.git   # FidelityFX fork needs SSH
cd RDNU
git -C ml/models/nss lfs pull                    # 15 MB of weights

Datasets are opt-in submodules; fetch instructions and licences: docs/Datasets.md. Windows build (VS Code + CMake, DXC): docs/build-windows.md.

Validating the network on the GPU

pip install torch numpy                          # CPU is fine
python runtime/tools/nss_export.py               # -> runtime/tools/golden/nss_backbone{,_int8}.rdnut
cmake --build build --target run_engine          # runs every bundle through the DX12 engine

run_engine prints max|err| per bundle and per checkpoint; fp32 must be within 1e-3, INT8 bit-exact against the integer-path reference.

Documents

Licences

RDNU is GPL-3.0 (LICENSE). Arm NSS weights: Arm AI Model Community License v1.0 (commercial use permitted; acceptable-use pass-through, notices, marked modifications). Model gym and capture plugin: Apache-2.0. Arm Neural Graphics SDK: MIT. RDG: Apache-2.0. Datasets: see docs/Datasets.md.

Acknowledgments

Arm (Neural Super Sampling, Model Gym, SDK), the RDG authors, and the creators of the GameIR, M3VIR, MPI-Sintel and Virtual KITTI 2 datasets.

About

Radeon Decoupled Neural Upscaler - RDNA3-focused ML-based uUpscaler using sunny2109/RDG (Decoupled G-buffer Guidance). Pretrained on guluthemonster/M3VIR and LLLebin/GameIR, fine-tuned on vKITTI2 and MPI-Sintel.

Resources

Stars

16 stars

Watchers

6 watching

Forks

Releases

Packages

Contributors

Languages