About

BioNeMo Inference Runtime (BioIR) is NVIDIA's library for structure-prediction inference. A five-stage GPU pipeline turns AlphaFold-lineage and all-atom models into PDB/mmCIF with confidence scores. Models stay ordinary nn.Modules — no TensorRT engine build.

Disclaimer

BioIR is in a Beta state, may not be fully functional, may contain errors or design flaws, and may be changed at any time without notice. We appreciate your feedback to improve and iterate on our products.

Getting Started

The notice and disclaimer applies to separately downloaded dependencies, model checkpoints, and metadata.

Prerequisites

• Linux, x86_64 or aarch64, with an NVIDIA GPU. The wheels are manylinux_2_34, so the host needs glibc 2.34 or newer — Ubuntu 22.04, RHEL 9 or later.
• Driver 580 or newer. The dev image carries a CUDA 13.2 build of PyTorch. An older driver runs it only through the forward-compatibility shim, which we have measured hanging and crashing part-way through a run rather than merely running slowly — results taken on one are discarded, not corrected.
• Python 3.12. The released wheels are tagged cp312, so pip finds no matching build on a newer interpreter.
• Docker and the NVIDIA Container Toolkit, to build from source in the dev container. Installing the wheel needs neither.

PyTorch and the CUDA math libraries arrive as wheel dependencies, or in nvcr.io/nvidia/pytorch:26.05-py3 when you use the container. Building the extension from source outside a container needs a C++17 compiler and CUDA headers as well — docs/dev.md.

Supported GPUs

BioIR runs on the Ampere, Ada Lovelace, Hopper, and Blackwell GPUs listed in the support matrix. The doc also shows which fused kernels apply to each architecture. Benchmarks page shows detailed speedup, memory, and accuracy for these GPUs: H200, H100, A100, L40S, GB200, and GB300.

Install

BioIR is published on PyPI, one wheel per CPU architecture:

pip install bionemo-ir

The wheel ships the kernels precompiled, so nothing in the install builds CUDA and running it needs only the driver's libcuda.so.1. That is the whole install if you are calling BioIR from your own code — the container below is for working on BioIR itself. docs/install.md covers the environment setup and the requirements in full.

Build from source

Configure SSH authentication with GitHub, then clone the repository and fetch its submodules and LFS objects.

git lfs install &&
  GIT_LFS_SKIP_SMUDGE=0 \
    git clone --recurse-submodules \
    git@github.com:NVIDIA-BioNeMo/BioNeMo-Inference-Runtime.git &&
  cd BioNeMo-Inference-Runtime

On a host with the required compiler and CUDA headers, uv provisions the pinned Python and exact dependency set, builds the editable extension, and runs tools inside that environment:

uv sync --locked
uv run --locked scripts/fetch_weights.sh --model boltz-2
uv run --locked python examples/folding/run_demo.py --output-dir output

The supported container workflow remains available when the host is not configured:

docker/dev.sh

Then run inside its shell:

uv lock --check
uv pip install --no-deps -e .
scripts/fetch_weights.sh --model boltz-2
python examples/folding/run_demo.py --output-dir output

Checkpoints come from their upstream publishers and need no NVIDIA credentials; anything that cannot be fetched is skipped, and the tests needing it skip too. Running uv run --locked scripts/run_tests.sh on the host, or scripts/run_tests.sh in the container, stages weights and runs the suite the way CI does.

See docs/dev.md for the host prerequisites, environment details and wheel build. docs/ref/docker-images.md covers the images and what docker/dev.sh mounts.

Documentation

BioIR documentation lives under docs/ and is published with Fern:

• Overview — what BioIR is and how to start
• Installation — requirements and release-wheel installation
• Quickstart — run a serial Boltz-2 prediction
• Ray multi-GPU inference — scale independent requests across visible GPUs
• Accelerate a custom model — swap BioIR modules into your own PyTorch model
• Port a data pipeline — run a new model family through build_processor
• Validate a ported pipeline — feature equivalence and end-to-end accuracy against upstream
• Developer guide — build, test, stage weights, contribute
• API reference — build_processor, model constructors, inputs/outputs
• Architecture — five-stage pipeline and runtime design
• Config architecture — model BaseConfig tree and pipeline stage configs
• Support matrix — models, GPUs, and fused kernels
• Token padding — what BioIR pads to a multiple of 8 tokens, and why it runs faster
• Benchmarks — measured speedup and memory against OSS PyTorch
• Model weights — checkpoint resolution and staging
• Docker images — development and runtime images
• Coding guidelines — style, naming, and tooling
• Folding example — runnable build_processor demo

Benchmarks

Methodology

Folding benchmarks over a bench set the shipped rebuild_dataset.py builds from RCSB and NVIDIA's MSA Search NIM — there is no dataset release to download. Template-bearing samples included: both sides load every bundled MSA and attach every listed template.

• One GPU, serial, one structure per forward call.
• Time only GPU-synchronized model.forward(). Featurization, transfers, postprocessing, writing, and scoring stay outside the window.
• Discard one warmup forward, then report one measured forward per sample.
• BioIR runs its default optimized config, which captures CUDA graphs of each model's trunk recycle, diffusion step, and confidence Pairformer (OpenFold2: trunk recycle only) for inputs up to 1024 tokens, or 2048 for the diffusion step on GPUs with at least 64 GiB; larger inputs run those regions eagerly. The results below were measured 2026-10-14.
• OSS runs its own inference script: eager always, plus torch.compile when it passes a dynamic-shape probe.
• Runtime knobs match on both sides — 200 sampling steps, 3 or 5 diffusion samples, and per-model recycling.
• Score written structures with OpenStructure lDDT and DockQ. Speedup is OSS forward / BioIR forward; above 1 favors BioIR.
• Future work will add additional Blackwell-optimized kernels.

Results

Geomean speedup, vs OSS torch.compile / vs OSS PyTorch eager; above 1 favours BioIR. Protenix has no torch.compile path. Twelve GPUs, per-model accuracy and peak memory, and how to reproduce any of it: docs/ref/benchmark.md.

The bench-perf-oss agent skill has the full gates, environment isolation, result schema, and charting protocol.

Contributing

We welcome contributions. See contributing.md for policy and docs/dev.md for the development workflow.

Citation

If you use BioIR in your research, please cite it via CITATION.cff.

Contact / Support

• Bugs and feature requests: GitHub Issues
• Usage questions: GitHub Discussions
• Security vulnerabilities: see SECURITY.md — do not file a public issue

License

NVIDIA-authored BioIR code is licensed under the Apache License 2.0. Distribution compliance material is available here:

• Third-party notices and attributions
• Full third-party license texts
• Gemmi 0.6.5 corresponding source, licensed under MPL-2.0 or LGPL-3.0-or-later; BioIR distributes it under the MPL-2.0 option

Notice and Disclaimer

The following notice applies to separately downloaded components, including runtime dependencies, model checkpoints, and chemical metadata.

NOTICE AND DISCLAIMER: This software automatically retrieves, accesses or interacts with external materials. Those retrieved materials are not distributed with this software and are governed solely by separate terms, conditions and licenses. You are solely responsible for finding, reviewing and complying with all applicable terms, conditions, and licenses, and for verifying the security, integrity and suitability of any retrieved materials for your specific use case. This software is provided "AS IS", without warranty of any kind. The author makes no representations or warranties regarding any retrieved materials, and assumes no liability for any losses, damages, liabilities or legal consequences from your use or inability to use this software or any retrieved materials. Use this software and the retrieved materials at your own risk.