v2.0.4

What's Changed

• Add optional inference-time diffusion steering (physical guidance) by @heya5 in #17
• Release 2.0.4: optional inference-time diffusion steering by @heya5 in #18

Full Changelog: v2.0.3...v2.0.4

v2.0.3 — IntelliFold-v2 on the AlphaFold 3 JAX engine

IntelliFold-v2 now runs on the AlphaFold 3 JAX engine for much faster inference — enabled by AF3's Apache-2.0 relicense in v3.0.3.

# if you already have alphafold3; otherwise build the vendored AF3 first (see README)
pip install intellifold

wget https://huggingface.co/intelligenAI/intellifold/resolve/main/fold_input.json
intellifold predict fold_input.json --model-dir=model_v2 --output-dir results -- --norun_data_pipeline

• Run IntelliFold-v2 on AF3's JAX stack; batch across all GPUs with --gpus all.
• Weights auto-download from Hugging Face on first run — no PyTorch needed.
• Vendored AF3 v3.0.3 under third_party/alphafold3/; packaging via pyproject.toml.

v2.0.2

v2.0.2 — Template Support

This release adds template-supported inference.

Key Changes

• Enable template features via --use_template
• Support both online template search and precomputed template inputs in YAML
• Documentation updates with new usage examples

Quick example

intellifold predict ./examples/5S8I_A_wtemplate.yaml --out_dir ./output --use_template

IntelliFold v2.0.0

Release IntelliFold v2.0.0

v0.1.0: IntelliFold Initial Version

Version 0.1.0