Commit ·
a40d48a
0
Parent(s):
Duplicate from nvidia/RFD4-Proteina
Browse filesCo-authored-by: Kieran Didi <kierandidi@users.noreply.huggingface.co>
This view is limited to 50 files because it contains too many changes. See raw diff
- .gitattributes +219 -0
- LICENSE +28 -0
- README.md +309 -0
- adapter/adapter_config.json +46 -0
- adapter/adapter_model.safetensors +3 -0
- ae_checkpoints/rfd4_proteina_ae-ema.ckpt +3 -0
- ae_checkpoints/rfd4_proteina_ae-ema.ckpt.sha256 +1 -0
- alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sif +3 -0
- alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sif.sha256 +1 -0
- alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sqsh +3 -0
- alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sqsh.sha256 +1 -0
- alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sif +3 -0
- alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sif.sha256 +1 -0
- alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sqsh +3 -0
- alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sqsh.sha256 +1 -0
- assets/design-gallery.png +3 -0
- assets/examples/antibody_design.png +3 -0
- assets/examples/enzyme_design.png +3 -0
- assets/examples/protein_binder.png +3 -0
- assets/wordmark-dark.webp +0 -0
- assets/wordmark.webp +0 -0
- base/config.json +263 -0
- base/model.safetensors +3 -0
- benchmarks/METHODS.md +56 -0
- benchmarks/provenance.json +13 -0
- benchmarks/source_data.csv +0 -0
- benchmarks/summary.csv +53 -0
- config.json +263 -0
- examples/antibody_design/design.cif +0 -0
- examples/antibody_design/input.cif +0 -0
- examples/enzyme_design/design.cif +0 -0
- examples/enzyme_design/input.cif +177 -0
- examples/monomer.cif +78 -0
- examples/protein_binder/design.cif +0 -0
- examples/protein_binder/input.cif +0 -0
- flow_checkpoints/rfd4_proteina_flow-ema.ckpt +3 -0
- flow_checkpoints/rfd4_proteina_flow-ema.ckpt.sha256 +1 -0
- flow_checkpoints/rfd4_proteina_flow_pretrain-ema.ckpt +3 -0
- flow_checkpoints/rfd4_proteina_flow_pretrain-ema.ckpt.sha256 +1 -0
- inference.yaml +79 -0
- model.safetensors +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda12.docker.tar +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda12.docker.tar.sha256 +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda12.sif +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda12.sif.sha256 +1 -0
- openfold3/runtime/of3_af3_refolding_env-cuda12.sqsh +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda12.sqsh.sha256 +1 -0
- openfold3/runtime/of3_af3_refolding_env-cuda13.docker.tar +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda13.docker.tar.sha256 +3 -0
- openfold3/runtime/of3_af3_refolding_env-cuda13.sif +3 -0
.gitattributes
ADDED
|
@@ -0,0 +1,219 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
| 2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
| 3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
| 4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
| 5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
| 6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
| 7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
| 8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
| 9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
| 10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
| 11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
| 12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
| 13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
| 14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
| 15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
| 16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
| 17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
| 18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
| 19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
| 20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
| 21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
| 22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
| 23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
| 24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
| 25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
| 26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
| 27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
| 28 |
+
*.tar filter=lfs diff=lfs merge=lfs -text
|
| 29 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
| 30 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
| 31 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
| 32 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
| 33 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
af3/apptainer_images/af3-open-36fad16.sif filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
aoti/h200/b1-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
aoti/h200/b2-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
aoti/h200/b4-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
aoti/h200/b8-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 41 |
+
aoti/h200/b1-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
aoti/h200/b1-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 43 |
+
aoti/h200/b1-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 44 |
+
aoti/h200/b1-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 45 |
+
aoti/blackwell/b1-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 46 |
+
aoti/blackwell/b2-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 47 |
+
aoti/blackwell/b4-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 48 |
+
aoti/blackwell/b8-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 49 |
+
aoti/blackwell/b1-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 50 |
+
aoti/blackwell/b1-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 51 |
+
aoti/pasco-blackwell/b1-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 52 |
+
aoti/pasco-blackwell/b1-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 53 |
+
aoti/pasco-blackwell/b1-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 54 |
+
aoti/pasco-blackwell/b1-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 55 |
+
aoti/pasco-blackwell/b1-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 56 |
+
aoti/pasco-blackwell/b1-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 57 |
+
aoti/pasco-blackwell/b1-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 58 |
+
aoti/pasco-blackwell/b1-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 59 |
+
aoti/pasco-blackwell/b1-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 60 |
+
aoti/pasco-blackwell/b1-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 61 |
+
aoti/pasco-blackwell/b1-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 62 |
+
aoti/pasco-blackwell/b1-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 63 |
+
aoti/pasco-blackwell/b1-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 64 |
+
aoti/pasco-blackwell/b1-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 65 |
+
aoti/pasco-blackwell/b1-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 66 |
+
aoti/pasco-blackwell/b2-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 67 |
+
aoti/pasco-blackwell/b4-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 68 |
+
aoti/pasco-blackwell/b8-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 69 |
+
aoti/pasco-blackwell/b2-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 70 |
+
aoti/pasco-blackwell/b4-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 71 |
+
aoti/pasco-blackwell/b8-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 72 |
+
aoti/pasco-blackwell/b2-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 73 |
+
aoti/pasco-blackwell/b4-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 74 |
+
aoti/pasco-blackwell/b8-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 75 |
+
aoti/pasco-blackwell/b2-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 76 |
+
aoti/pasco-blackwell/b4-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 77 |
+
aoti/pasco-blackwell/b8-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 78 |
+
aoti/pasco-blackwell/b2-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 79 |
+
aoti/pasco-blackwell/b4-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 80 |
+
aoti/pasco-blackwell/b8-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 81 |
+
openfold3/runtime/openfold3-36fad16.sif filter=lfs diff=lfs merge=lfs -text
|
| 82 |
+
assets/examples/protein_binder.png filter=lfs diff=lfs merge=lfs -text
|
| 83 |
+
assets/examples/enzyme_design.png filter=lfs diff=lfs merge=lfs -text
|
| 84 |
+
assets/examples/antibody_design.png filter=lfs diff=lfs merge=lfs -text
|
| 85 |
+
aoti/l40s/b1-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 86 |
+
aoti/l40s/b1-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 87 |
+
aoti/l40s/b1-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 88 |
+
aoti/l40s/b1-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 89 |
+
aoti/l40s/b1-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 90 |
+
aoti/l40s/b1-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 91 |
+
aoti/l40s/b1-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 92 |
+
aoti/l40s/b1-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 93 |
+
aoti/a100/b1-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 94 |
+
aoti/a100/b1-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 95 |
+
aoti/a100/b1-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 96 |
+
aoti/a100/b1-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 97 |
+
aoti/a100/b1-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 98 |
+
aoti/a100/b1-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 99 |
+
aoti/a100/b1-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 100 |
+
aoti/a100/b1-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 101 |
+
aoti/l40s/b1-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 102 |
+
aoti/l40s/b1-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 103 |
+
aoti/l40s/b1-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 104 |
+
aoti/l40s/b1-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 105 |
+
aoti/l40s/b1-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 106 |
+
aoti/l40s/b1-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 107 |
+
aoti/l40s/b1-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 108 |
+
aoti/l40s/b2-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 109 |
+
aoti/l40s/b2-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 110 |
+
aoti/l40s/b2-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 111 |
+
aoti/l40s/b2-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 112 |
+
aoti/l40s/b2-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 113 |
+
aoti/l40s/b2-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 114 |
+
aoti/l40s/b2-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 115 |
+
aoti/l40s/b2-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 116 |
+
aoti/l40s/b2-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 117 |
+
aoti/l40s/b2-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 118 |
+
aoti/l40s/b2-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 119 |
+
aoti/l40s/b2-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 120 |
+
aoti/l40s/b2-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 121 |
+
aoti/l40s/b2-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 122 |
+
aoti/l40s/b2-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 123 |
+
aoti/l40s/b4-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 124 |
+
aoti/l40s/b4-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 125 |
+
aoti/l40s/b4-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 126 |
+
aoti/l40s/b4-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 127 |
+
aoti/l40s/b4-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 128 |
+
aoti/l40s/b4-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 129 |
+
aoti/l40s/b4-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 130 |
+
aoti/l40s/b4-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 131 |
+
aoti/l40s/b4-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 132 |
+
aoti/l40s/b4-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 133 |
+
aoti/l40s/b4-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 134 |
+
aoti/l40s/b4-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 135 |
+
aoti/l40s/b4-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 136 |
+
aoti/l40s/b4-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 137 |
+
aoti/l40s/b4-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 138 |
+
aoti/l40s/b8-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 139 |
+
aoti/l40s/b8-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 140 |
+
aoti/l40s/b8-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 141 |
+
aoti/l40s/b8-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 142 |
+
aoti/l40s/b8-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 143 |
+
aoti/l40s/b8-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 144 |
+
aoti/l40s/b8-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 145 |
+
aoti/l40s/b8-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 146 |
+
aoti/l40s/b8-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 147 |
+
aoti/l40s/b8-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 148 |
+
aoti/l40s/b8-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 149 |
+
aoti/l40s/b8-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 150 |
+
aoti/l40s/b8-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 151 |
+
aoti/a100/b1-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 152 |
+
aoti/a100/b1-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 153 |
+
aoti/a100/b1-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 154 |
+
aoti/a100/b1-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 155 |
+
aoti/a100/b1-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 156 |
+
aoti/a100/b1-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 157 |
+
aoti/a100/b1-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 158 |
+
aoti/a100/b2-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 159 |
+
aoti/a100/b2-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 160 |
+
aoti/a100/b2-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 161 |
+
aoti/a100/b2-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 162 |
+
aoti/a100/b2-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 163 |
+
aoti/a100/b2-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 164 |
+
aoti/a100/b2-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 165 |
+
aoti/a100/b2-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 166 |
+
aoti/a100/b2-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 167 |
+
aoti/a100/b2-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 168 |
+
aoti/a100/b2-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 169 |
+
aoti/a100/b2-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 170 |
+
aoti/a100/b2-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 171 |
+
aoti/a100/b2-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 172 |
+
aoti/a100/b2-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 173 |
+
aoti/a100/b4-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 174 |
+
aoti/a100/b4-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 175 |
+
aoti/a100/b4-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 176 |
+
aoti/a100/b4-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 177 |
+
aoti/a100/b4-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 178 |
+
aoti/a100/b4-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 179 |
+
aoti/a100/b4-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 180 |
+
aoti/a100/b4-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 181 |
+
aoti/a100/b4-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 182 |
+
aoti/a100/b4-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 183 |
+
aoti/a100/b4-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 184 |
+
aoti/a100/b4-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 185 |
+
aoti/a100/b4-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 186 |
+
aoti/a100/b4-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 187 |
+
aoti/a100/b4-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 188 |
+
aoti/a100/b8-n1024/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 189 |
+
aoti/a100/b8-n128/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 190 |
+
aoti/a100/b8-n160/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 191 |
+
aoti/a100/b8-n192/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 192 |
+
aoti/a100/b8-n224/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 193 |
+
aoti/a100/b8-n256/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 194 |
+
aoti/a100/b8-n288/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 195 |
+
aoti/a100/b8-n320/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 196 |
+
aoti/a100/b8-n352/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 197 |
+
aoti/a100/b8-n384/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 198 |
+
aoti/a100/b8-n448/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 199 |
+
aoti/a100/b8-n512/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 200 |
+
aoti/a100/b8-n640/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 201 |
+
aoti/a100/b8-n768/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 202 |
+
aoti/a100/b8-n896/root/package.pt2 filter=lfs diff=lfs merge=lfs -text
|
| 203 |
+
assets/design-gallery.png filter=lfs diff=lfs merge=lfs -text
|
| 204 |
+
openfold3/runtime/of3_af3_refolding_env.sif filter=lfs diff=lfs merge=lfs -text
|
| 205 |
+
alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sif filter=lfs diff=lfs merge=lfs -text
|
| 206 |
+
alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 207 |
+
openfold3/runtime/of3_af3_refolding_env.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 208 |
+
openfold3/runtime/of3_af3_refolding_env-cuda12.sif filter=lfs diff=lfs merge=lfs -text
|
| 209 |
+
openfold3/runtime/of3_af3_refolding_env-cuda12.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 210 |
+
alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sif filter=lfs diff=lfs merge=lfs -text
|
| 211 |
+
alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 212 |
+
openfold3/runtime/of3_af3_refolding_env-cuda13.sif filter=lfs diff=lfs merge=lfs -text
|
| 213 |
+
openfold3/runtime/of3_af3_refolding_env-cuda13.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 214 |
+
runtime/rfd4-proteina-cpu-amd64.sif filter=lfs diff=lfs merge=lfs -text
|
| 215 |
+
runtime/rfd4-proteina-cpu-amd64.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 216 |
+
runtime/rfd4-proteina-cu12-amd64.sif filter=lfs diff=lfs merge=lfs -text
|
| 217 |
+
runtime/rfd4-proteina-cu12-amd64.sqsh filter=lfs diff=lfs merge=lfs -text
|
| 218 |
+
runtime/rfd4-proteina-cu13-amd64.sif filter=lfs diff=lfs merge=lfs -text
|
| 219 |
+
runtime/rfd4-proteina-cu13-amd64.sqsh filter=lfs diff=lfs merge=lfs -text
|
LICENSE
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
BSD 3-Clause License
|
| 2 |
+
|
| 3 |
+
Copyright (c) 2026, Institute for Protein Design, University of Washington
|
| 4 |
+
|
| 5 |
+
Redistribution and use in source and binary forms, with or without
|
| 6 |
+
modification, are permitted provided that the following conditions are met:
|
| 7 |
+
|
| 8 |
+
* Redistributions of source code must retain the above copyright notice, this
|
| 9 |
+
list of conditions and the following disclaimer.
|
| 10 |
+
|
| 11 |
+
* Redistributions in binary form must reproduce the above copyright notice,
|
| 12 |
+
this list of conditions and the following disclaimer in the documentation
|
| 13 |
+
and/or other materials provided with the distribution.
|
| 14 |
+
|
| 15 |
+
* Neither the name of the copyright holder nor the names of its
|
| 16 |
+
contributors may be used to endorse or promote products derived from
|
| 17 |
+
this software without specific prior written permission.
|
| 18 |
+
|
| 19 |
+
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
| 20 |
+
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
| 21 |
+
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
|
| 22 |
+
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
|
| 23 |
+
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
| 24 |
+
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
| 25 |
+
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
| 26 |
+
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
|
| 27 |
+
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
| 28 |
+
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
README.md
ADDED
|
@@ -0,0 +1,309 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: bsd-3-clause
|
| 3 |
+
library_name: rfd4-proteina
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
datasets:
|
| 7 |
+
- nvidia/RFD4-Proteina
|
| 8 |
+
tags:
|
| 9 |
+
- biology
|
| 10 |
+
- protein-design
|
| 11 |
+
- biomolecular-design
|
| 12 |
+
- structure-generation
|
| 13 |
+
- flow-matching
|
| 14 |
+
- annotated-cif
|
| 15 |
+
---
|
| 16 |
+
<p align="center"><picture><source media="(prefers-color-scheme: dark)" srcset="assets/wordmark-dark.webp"><img src="assets/wordmark.webp" width="520" alt="RFD4-Proteína"></picture></p>
|
| 17 |
+
<p align="center"><strong>Sequence and structure, designed together.</strong><br>Baker Lab × NVIDIA GenAIR</p>
|
| 18 |
+
<p align="center"><a href="https://rfd4-proteina.org/">Project page</a> · <a href="https://github.com/RosettaCommons/RFD4-Proteina">Code</a> · <a href="https://rfd4-proteina.org/latest/">Documentation</a> · <a href="https://huggingface.co/datasets/nvidia/RFD4-Proteina">Training data</a> · <a href="https://huggingface.co/spaces/nvidia/rfd4-proteina-zerogpu">Studio (ZeroGPU)</a> · <a href="https://huggingface.co/spaces/nvidia/rfd4-proteina-space">Studio (dedicated GPU)</a></p>
|
| 19 |
+
|
| 20 |
+
# RFD4-Proteína
|
| 21 |
+
|
| 22 |
+
RFD4-Proteína is a flow-matching model for biomolecular design. It generates protein sequences and all-atom structures together, conditioned on the molecular context you want to preserve: a binding target, an enzyme motif, a ligand, or selected structural features.
|
| 23 |
+
|
| 24 |
+
<img src="assets/design-gallery.png" width="1050" alt="Examples from the project: nanobody complex, protein–DNA binder, RNA pseudoknot, ligand pocket, glycan complex, and protein assembly.">
|
| 25 |
+
|
| 26 |
+
## Where to find each entry point
|
| 27 |
+
|
| 28 |
+
This repository is the **canonical native bundle**. `config.json` and `model.safetensors` at
|
| 29 |
+
the root are the *native* weights, loaded by `Proteina.from_pretrained`.
|
| 30 |
+
|
| 31 |
+
| Call | Weights | Location | Status |
|
| 32 |
+
|---|---|---|---|
|
| 33 |
+
| `Proteina.from_pretrained("nvidia/RFD4-Proteina")` | Native | root of this repo | **Supported** |
|
| 34 |
+
| `RFD4ProteinaPipeline.from_pretrained("nvidia/RFD4-Proteina-Diffusers")` | Diffusers | [its own repo](https://huggingface.co/nvidia/RFD4-Proteina-Diffusers) | **Supported**, Diffusers from git until the next release |
|
| 35 |
+
|
| 36 |
+
It is one model throughout — nothing is retrained. The Diffusers weights are the same tensors
|
| 37 |
+
re-prefixed and split across `transformer/` and `vae/`. Diffusers needs `model_index.json` at the
|
| 38 |
+
root of whatever it is pointed at, so it has its own repository. Annotated-CIF preprocessing for
|
| 39 |
+
Diffusers is in [`nvidia/RFD4-Proteina-prep`](https://huggingface.co/nvidia/RFD4-Proteina-prep).
|
| 40 |
+
|
| 41 |
+
> **Diffusers from git until the next release.** `nvidia/RFD4-Proteina-Diffusers` holds the same 2026-10-19
|
| 42 |
+
> release. `RFD4ProteinaPipeline` was merged into Diffusers on 2026-10-20 ([huggingface/diffusers#15088](https://github.com/huggingface/diffusers/pull/15088)) and ships in the next
|
| 43 |
+
> release; until then install Diffusers from GitHub, as in the snippet below.
|
| 44 |
+
|
| 45 |
+
Training data lives in the companion [dataset repository](https://huggingface.co/datasets/nvidia/RFD4-Proteina).
|
| 46 |
+
|
| 47 |
+
## At a glance
|
| 48 |
+
|
| 49 |
+
| | Details |
|
| 50 |
+
|---|---|
|
| 51 |
+
| **Model** | Joint sequence–structure generation with a matched flow model and autoencoder |
|
| 52 |
+
| **Inputs** | Annotated CIF structures or YAML condition specifications |
|
| 53 |
+
| **Outputs** | Designed molecular structures in CIF format, with generated residue identities and coordinates |
|
| 54 |
+
| **Weights** | EMA weights; 435,622,465 stored parameters across the paired bundle, in FP32 |
|
| 55 |
+
| **Download** | 1.74 GB for `model.safetensors`, plus `config.json`; training data is a separate download |
|
| 56 |
+
| **Runtime** | RFD4 Python package; Linux x86_64, Python 3.12, supported NVIDIA GPU and CUDA 13-compatible driver |
|
| 57 |
+
| **License** | [BSD 3-Clause](LICENSE) for the RFD4 model and weights; optional third-party assets have separate terms |
|
| 58 |
+
|
| 59 |
+
The bundle includes both model components. Use the RFD4 loader below; this is a native RFD4 bundle.
|
| 60 |
+
|
| 61 |
+
## Checkpoints in this release
|
| 62 |
+
|
| 63 |
+
The release is a **matched pair**: a flow model and the autoencoder it was trained against. They are not interchangeable across releases — the autoencoder defines the latent space the flow model generates in, and mixing a flow model with a different autoencoder produces nonsense rather than an error.
|
| 64 |
+
|
| 65 |
+
| | Current release (2026-10-19) | Previous release |
|
| 66 |
+
|---|---|---|
|
| 67 |
+
| Flow model | `rfd4-proteina-flow-release-2026-10-19`: the pretrained base (step 320000) with a rank-32 LoRA fine-tune merged in | `chk_epoch_00000169_step_000000340000-ema` |
|
| 68 |
+
| Autoencoder | `rfd4-proteina-ae-release-2026-10-19` | `2026_09_04_release_ae_z16_400k-ema` |
|
| 69 |
+
| Latent dimension | 32 | 16 |
|
| 70 |
+
| Residue vocabulary | 147 classes | 33 classes |
|
| 71 |
+
| Stored parameters | 435,622,465 | 434,348,737 |
|
| 72 |
+
|
| 73 |
+
The 2026-10-19 pair supersedes the 2026-10-15 pair (`2026_10_15_hero_s43_c512_80k-ema` and `2026_10_15_unified_ae_s43_400k-ema`), which has the same architecture. The pretrained base alone is in `base/`, and the LoRA adapter that turns it into the release model is in `adapter/` (PEFT format, rank 32 on `to_qkv` and `to_out_node`); the release bundle at the root is exactly the base plus the adapter, merged.
|
| 74 |
+
|
| 75 |
+
The current release's autoencoder is **unified**: one autoencoder covering proteins, nucleic acids and small molecules, with a 147-class residue vocabulary in place of the previous 33-class protein vocabulary, and a 32-dimensional latent in place of 16. This is the change that makes the two releases incompatible with each other's components.
|
| 76 |
+
|
| 77 |
+
| File | Use |
|
| 78 |
+
|---|---|
|
| 79 |
+
| [`config.json`](config.json) + `model.safetensors` | Recommended paired bundle for `Proteina.from_pretrained` |
|
| 80 |
+
| `base/` and `adapter/` | The pretrained base bundle and the PEFT adapter that makes it the release model (`adapter_path=` in the native runtime) |
|
| 81 |
+
| `flow_checkpoints/rfd4_proteina_flow-ema.ckpt` | Native Lightning flow EMA weights |
|
| 82 |
+
| `ae_checkpoints/rfd4_proteina_ae-ema.ckpt` | Matching native Lightning autoencoder EMA weights |
|
| 83 |
+
| [`release-checkpoints.json`](release-checkpoints.json) | Pair identity, stable aliases, source filenames, and SHA-256 checksums |
|
| 84 |
+
| [`provenance.json`](provenance.json) | Source archive and original checkpoint provenance |
|
| 85 |
+
|
| 86 |
+
The published `.ckpt` files carry **EMA weights and hyperparameters only**. The duplicate `SimpleEMA` shadow copy, optimizer, scheduler and loop state are removed, which is why they are 1.74 GB and 1.03 GB rather than 4.21 GB and 3.10 GB. Their weight tensors are bit-identical to the full training checkpoints, whose SHA-256 values are recorded in `release-checkpoints.json`. They initialize weights for inference or fine-tuning; **resuming an interrupted training run needs the full checkpoint**, which is not distributed here.
|
| 87 |
+
|
| 88 |
+
Stable filenames always select the current release; the dated training filename of each is recorded in `release-checkpoints.json`. Pin a Hub revision and verify the SHA-256 sidecars for exact identity. These EMA exports initialize weights; resuming optimizer and scheduler state requires a full training checkpoint.
|
| 89 |
+
|
| 90 |
+
## Start with one design
|
| 91 |
+
|
| 92 |
+
### 1. Install
|
| 93 |
+
|
| 94 |
+
Pick one route; the [quickstart](https://rfd4-proteina.org/latest/quickstart.html) covers all three in detail.
|
| 95 |
+
|
| 96 |
+
**GitHub** (recommended: the full workflow with annotated inputs, reward-guided search and refolding metrics):
|
| 97 |
+
|
| 98 |
+
```bash
|
| 99 |
+
git clone https://github.com/RosettaCommons/RFD4-Proteina.git
|
| 100 |
+
cd RFD4-Proteina
|
| 101 |
+
make setup
|
| 102 |
+
source .venv/bin/activate
|
| 103 |
+
```
|
| 104 |
+
|
| 105 |
+
**PyPI** (the model inside your own Python environment):
|
| 106 |
+
|
| 107 |
+
```bash
|
| 108 |
+
python3.12 -m venv .venv && source .venv/bin/activate
|
| 109 |
+
pip install --extra-index-url https://download.pytorch.org/whl/cu132 "rfd4-proteina[metrics-cuda13]"
|
| 110 |
+
```
|
| 111 |
+
|
| 112 |
+
**Diffusers:** see [Loading the model](#diffusers) below.
|
| 113 |
+
|
| 114 |
+
**Container images** (no build needed): prebuilt RFD4-Proteína runtime images for Linux x86_64 are in [`runtime/`](runtime), as Apptainer (`.sif`) and SquashFS for Enroot/Pyxis (`.sqsh`) files with SHA-256 sidecars. `cu13` (CUDA 13), `cu12` (CUDA 12) and `cpu` (no GPU; imports, CLI and CPU metrics only) variants exist, built from the public [Apptainer recipe](https://github.com/RosettaCommons/RFD4-Proteina/blob/main/apptainer/rfd4-proteina.def). They contain the code and its dependencies, not model weights or the AF2/AF3 oracles; pass `hf_ckpt_path=nvidia/RFD4-Proteina` and a writable `HF_HOME`.
|
| 115 |
+
|
| 116 |
+
```bash
|
| 117 |
+
hf download nvidia/RFD4-Proteina runtime/rfd4-proteina-cu13-amd64.sif runtime/rfd4-proteina-cu13-amd64.sif.sha256 --local-dir .
|
| 118 |
+
apptainer exec --nv runtime/rfd4-proteina-cu13-amd64.sif rfd4-generate hf_ckpt_path=nvidia/RFD4-Proteina inputs=in.cif out_dir=./runs
|
| 119 |
+
```
|
| 120 |
+
|
| 121 |
+
Setup caches the paired model bundle; downloads use `HF_HOME` when set. You do not need the training dataset for generation. If access to this repository is restricted, run `hf auth login` first. See [installation](https://rfd4-proteina.org/latest/installation.html) for hardware and build requirements.
|
| 122 |
+
|
| 123 |
+
### 2. Generate a small protein
|
| 124 |
+
|
| 125 |
+
From the checkout root (with the PyPI install, use the packaged example instead: `inputs=$(python -c "import rfproteina, pathlib; print(pathlib.Path(rfproteina.__file__).parent / 'data/benchmarks/monomer/monomer-short.cif')")`):
|
| 126 |
+
|
| 127 |
+
```bash
|
| 128 |
+
rfd4-generate inputs=data/benchmarks/monomer/monomer-short.cif \
|
| 129 |
+
hf_ckpt_path=nvidia/RFD4-Proteina \
|
| 130 |
+
hf_ckpt_revision=a06e2509c7033b010fc43614e9fd840303588502 \
|
| 131 |
+
num_replicates=1 out_dir=./runs/first '~metrics' compile_lengths=null
|
| 132 |
+
```
|
| 133 |
+
|
| 134 |
+
Find the generated CIF under `runs/first/inference_samples/`. This first run disables external scoring and compilation to keep setup simple. For repeated runs, use the normal compiled configuration described in [running the model](https://rfd4-proteina.org/latest/how-to/run-the-model.html); compilation has an initial warmup cost.
|
| 135 |
+
|
| 136 |
+
### 3. Design a PD-L1 binder or an enzyme scaffold
|
| 137 |
+
|
| 138 |
+
```bash
|
| 139 |
+
hf download nvidia/RFD4-Proteina \
|
| 140 |
+
--revision a06e2509c7033b010fc43614e9fd840303588502 \
|
| 141 |
+
--local-dir ./rfd4-examples \
|
| 142 |
+
examples/protein_binder/input.cif examples/enzyme_design/input.cif
|
| 143 |
+
|
| 144 |
+
rfd4-generate inputs=./rfd4-examples/examples/protein_binder/input.cif \
|
| 145 |
+
hf_ckpt_path=nvidia/RFD4-Proteina \
|
| 146 |
+
hf_ckpt_revision=a06e2509c7033b010fc43614e9fd840303588502 \
|
| 147 |
+
num_replicates=1 out_dir=./runs/pdl1 '~metrics' compile_lengths=null
|
| 148 |
+
```
|
| 149 |
+
|
| 150 |
+
For enzyme motif scaffolding, replace the input with `./rfd4-examples/examples/enzyme_design/input.cif` and choose a new output directory. The annotations specify what to hold fixed and what to design; an ordinary, unannotated structure is not a complete design specification.
|
| 151 |
+
|
| 152 |
+
| Design goal | Start here |
|
| 153 |
+
|---|---|
|
| 154 |
+
| Protein–protein binder | [PD-L1 input](examples/protein_binder/input.cif) · [example output](examples/protein_binder/design.cif) |
|
| 155 |
+
| Enzyme motif scaffolding | [Enzyme input](examples/enzyme_design/input.cif) · [example output](examples/enzyme_design/design.cif) |
|
| 156 |
+
| Antibody design | [HER2 input](examples/antibody_design/input.cif) · [example output](examples/antibody_design/design.cif) |
|
| 157 |
+
| Unconditional monomer | [100-residue spec](examples/monomer.cif) — a length specification, no input coordinates |
|
| 158 |
+
| Prepare your own conditions | [Annotated CIFs and Python input preparation](https://rfd4-proteina.org/latest/how-to/input_prep/index.html) · [YAML specifications](https://rfd4-proteina.org/latest/how-to/input_prep/yaml-json.html) |
|
| 159 |
+
| Interactive design | [AtomWorks Studio](https://huggingface.co/spaces/nvidia/rfd4-proteina-space) on a dedicated GPU pool · [the same Studio on ZeroGPU](https://huggingface.co/spaces/nvidia/rfd4-proteina-zerogpu) if you would rather spend your own Hugging Face quota. Both screen every input and design with SecureDNA and refuse flagged requests. |
|
| 160 |
+
|
| 161 |
+
## Loading the model
|
| 162 |
+
|
| 163 |
+
### Native RFD4 (this repository)
|
| 164 |
+
|
| 165 |
+
Load the matched model and autoencoder once and reuse them across runs:
|
| 166 |
+
|
| 167 |
+
```python
|
| 168 |
+
from pathlib import Path
|
| 169 |
+
from huggingface_hub import hf_hub_download
|
| 170 |
+
from hydra import compose, initialize_config_dir
|
| 171 |
+
import rootutils
|
| 172 |
+
import rfproteina
|
| 173 |
+
from rfproteina.proteina import Proteina
|
| 174 |
+
from rfproteina.generate import run_generation
|
| 175 |
+
|
| 176 |
+
revision = "a06e2509c7033b010fc43614e9fd840303588502"
|
| 177 |
+
input_cif = hf_hub_download(
|
| 178 |
+
"nvidia/RFD4-Proteina", "examples/enzyme_design/input.cif", revision=revision,
|
| 179 |
+
)
|
| 180 |
+
model = Proteina.from_pretrained("nvidia/RFD4-Proteina", revision=revision)
|
| 181 |
+
config_dir = rootutils.find_root(rfproteina.__file__, indicator=".project-root") / "configs"
|
| 182 |
+
with initialize_config_dir(config_dir=str(config_dir), version_base="1.3"):
|
| 183 |
+
cfg = compose(config_name="inference", overrides=[
|
| 184 |
+
f"inputs={input_cif}", "out_dir=./runs/enzyme",
|
| 185 |
+
"num_replicates=1", "~metrics", "compile_lengths=null",
|
| 186 |
+
])
|
| 187 |
+
run_generation(cfg, model=model)
|
| 188 |
+
print(list(Path("runs/enzyme/inference_samples").rglob("*_generated.cif*")))
|
| 189 |
+
```
|
| 190 |
+
|
| 191 |
+
`from_pretrained` loads in evaluation mode on CPU by default; `run_generation` uses the native CUDA inference workflow. A local exported bundle directory also works as the model argument.
|
| 192 |
+
|
| 193 |
+
### Diffusers
|
| 194 |
+
|
| 195 |
+
```bash
|
| 196 |
+
# RFD4-Proteina was merged into Diffusers on 2026-10-20 and ships in the next release.
|
| 197 |
+
# Until that release, install Diffusers from git:
|
| 198 |
+
pip install git+https://github.com/huggingface/diffusers git+https://huggingface.co/nvidia/RFD4-Proteina-prep@c9959efdebd75ac7fe9129fbbf0e291a6e0487d5 accelerate
|
| 199 |
+
|
| 200 |
+
# Once the release is out, this is enough:
|
| 201 |
+
# pip install diffusers git+https://huggingface.co/nvidia/RFD4-Proteina-prep@c9959efdebd75ac7fe9129fbbf0e291a6e0487d5
|
| 202 |
+
```
|
| 203 |
+
|
| 204 |
+
```python
|
| 205 |
+
import torch
|
| 206 |
+
from diffusers import RFD4ProteinaPipeline
|
| 207 |
+
from rfd4_proteina_prep import prepare_cif, write_cif
|
| 208 |
+
|
| 209 |
+
pipe = RFD4ProteinaPipeline.from_pretrained("nvidia/RFD4-Proteina-Diffusers", dtype=torch.bfloat16).to("cuda:0")
|
| 210 |
+
features, atoms = prepare_cif("input.cif", seed=42, device=pipe.device)
|
| 211 |
+
required = ["padding_mask", "is_in_context_conditioning", "supports_local_latent", "token_type",
|
| 212 |
+
"atom37_mask", "coords_nm"]
|
| 213 |
+
conditioning = {name: value for name, value in features.items() if name not in required}
|
| 214 |
+
output = pipe(
|
| 215 |
+
padding_mask=features["padding_mask"],
|
| 216 |
+
is_in_context_conditioning=features["is_in_context_conditioning"],
|
| 217 |
+
supports_local_latent=features["supports_local_latent"],
|
| 218 |
+
token_type=features["token_type"],
|
| 219 |
+
atom37_mask=features["atom37_mask"],
|
| 220 |
+
conditioning_features=conditioning,
|
| 221 |
+
generator=torch.Generator("cuda:0").manual_seed(42),
|
| 222 |
+
)
|
| 223 |
+
write_cif(output, features, atoms, "design.cif")
|
| 224 |
+
```
|
| 225 |
+
|
| 226 |
+
This needs Diffusers installed from git, as shown above. Partial diffusion and inverse folding work in this port. See the [repository](https://huggingface.co/nvidia/RFD4-Proteina-Diffusers) for details.
|
| 227 |
+
|
| 228 |
+
### Compute presets
|
| 229 |
+
|
| 230 |
+
The released recipes differ only in how often the trunk is recycled per sampling step, as `(steps, recycle depth)` bands over 60 steps. `thinking_low|medium|high` are the front-loaded schedules, and the default is medium. The native runtime also has flat and mid-peak schedules for each budget; see the [running guide](https://rfd4-proteina.org/latest/how-to/run-the-model.html#recycling-schedules) for which one suits a task.
|
| 231 |
+
|
| 232 |
+
| Preset | Bands | Trunk passes |
|
| 233 |
+
|---|---|---:|
|
| 234 |
+
| Low | `[[10,4],[10,3],[10,2],[30,1]]` | 120 |
|
| 235 |
+
| **Medium (default)** | `[[5,6],[10,5],[5,4],[10,3],[20,2],[10,1]]` | **180** |
|
| 236 |
+
| High | `[[10,9],[10,8],[10,5],[20,3],[10,2]]` | 300 |
|
| 237 |
+
|
| 238 |
+
In the native runtime select one with `recycle_strategy=thinking_low|thinking_medium|thinking_high`.
|
| 239 |
+
|
| 240 |
+
## Hardware requirements
|
| 241 |
+
|
| 242 |
+
Tested and confirmed on the following NVIDIA GPUs. Linux x86_64, Python 3.12, and a CUDA 13-compatible driver are required.
|
| 243 |
+
|
| 244 |
+
| Architecture | Data center | Workstation and desktop |
|
| 245 |
+
|---|---|---|
|
| 246 |
+
| Blackwell | B200, B300, GB200, GB300, RTX PRO 6000 Blackwell Server Edition | RTX PRO 6000 Blackwell Workstation Edition, GB10 (DGX Spark) |
|
| 247 |
+
| Hopper | H100 SXM, H100 PCIe, H100 NVL, H200, H200 NVL, GH200 144GB | |
|
| 248 |
+
| Ada Lovelace | L40S | RTX 6000 Ada Generation |
|
| 249 |
+
| Ampere | A100 SXM4 40GB and 80GB, A100 PCIe 40GB and 80GB | |
|
| 250 |
+
|
| 251 |
+
Weights are 1.74 GB in FP32. Peak GPU memory depends on token count, batch size and recycle depth rather than on the weights; the quickstart above runs at batch size 1 with compilation disabled. Refolding and metrics need the selected oracle's own environment and weights, which are a separate download.
|
| 252 |
+
|
| 253 |
+
## Measured design efficiency
|
| 254 |
+
|
| 255 |
+
> **These numbers were measured on the previous release pair**, in the October 9, 2026 frozen project benchmark, which predates the current checkpoints. They have not been re-measured on the current release and should not be read as describing it.
|
| 256 |
+
|
| 257 |
+
The table uses one consistent high-thinking setting for RFD4 and the strongest non-RFD4 comparator in each family. **Metric:** mean per-target structurally distinct computational passers per GPU-hour, including generation and common AF3 evaluation. Higher is better.
|
| 258 |
+
|
| 259 |
+
| Benchmark | Targets | RFD4-Proteína, high thinking | Best other method in this comparison |
|
| 260 |
+
|---|---:|---:|---|
|
| 261 |
+
| Protein binders | 24 | **71.60** | Proteina-Complexa · 13.24 |
|
| 262 |
+
| Small-molecule binders | 77 | **112.25** | BoltzGen · 15.48 |
|
| 263 |
+
| Atomic Motif Enzyme | 37 | **91.18** | AlphaProtein-novo (Codesign) · 26.02 |
|
| 264 |
+
| DNA binders | 3 | **41.97** | BoltzGen (fixed DNA) · 9.46 |
|
| 265 |
+
|
| 266 |
+
There are 200 original attempts per target. Failed attempts consume time; setup, compilation, CPU scoring, and idle allocation time are excluded. Values are means of target-level rates, not pooled totals or single-design latency.
|
| 267 |
+
|
| 268 |
+
Conditioning differs between some methods: PPI uses RFD4 backbone conditioning versus all-atom conditioning for several baselines; DNA compares unfixed RFD4 inputs with fixed-DNA baselines; Proteina-Complexa's small-molecule route uses fixed ligands. Only three DNA targets are included. Computational passers do not establish experimental activity.
|
| 269 |
+
|
| 270 |
+
[Benchmark provenance](benchmarks/provenance.json) · [Full comparison table](benchmarks/summary.csv) · [Per-target source data](benchmarks/source_data.csv) · [Benchmark methods and pass criteria](benchmarks/METHODS.md)
|
| 271 |
+
|
| 272 |
+
## Training data and limitations
|
| 273 |
+
|
| 274 |
+
The companion [dataset card](https://huggingface.co/datasets/nvidia/RFD4-Proteina) describes the available training snapshot, component-specific filters, source terms, and loading format. Its upload date or proposed sampling mix should not be interpreted as the exact training history of every checkpoint.
|
| 275 |
+
|
| 276 |
+
Use the model for computational biomolecular design and follow designs with appropriate structural and experimental evaluation. Performance depends on the target, input annotations, sampling settings, and training coverage. Generation-only quickstarts omit external refolding and scoring; see [evaluation and metrics](https://rfd4-proteina.org/latest/how-to/evaluate-with-metrics.html) to add them.
|
| 277 |
+
|
| 278 |
+
<details>
|
| 279 |
+
<summary><strong>Other repository files</strong></summary>
|
| 280 |
+
|
| 281 |
+
**Third-party software and weights.** The BSD 3-Clause license covers the RFD4-Proteína model, weights and code only. The optional `alphafold2/runtime/` and `openfold3/runtime/` container images package third-party software (AlphaFold2- and AlphaFold3-derived code, OpenFold3 and their dependencies) that stays under its own upstream licenses, some of which restrict commercial use or redistribution. The ported OpenFold3 weights in `openfold3/weights/` are Apache-2.0 (see the provenance file). Review each upstream license before using the images, and do not assume the BSD license applies to them. Neither oracle is needed for generation.
|
| 282 |
+
|
| 283 |
+
Optional [AF3/OpenFold3 runtime images](openfold3/runtime) are available for Linux x86_64 with CUDA 12 or CUDA 13 in Apptainer (`.sif`), SquashFS (`.sqsh`) and Docker (`.docker.tar`) formats, with SHA-256 checksums. [OpenFold3 weights](openfold3/weights) are downloaded separately. These files are only needed for refolding and metrics; generation does not use them.
|
| 284 |
+
|
| 285 |
+
</details>
|
| 286 |
+
|
| 287 |
+
## License
|
| 288 |
+
|
| 289 |
+
The RFD4-Proteína model and weights are released under the [BSD 3-Clause License](LICENSE), Copyright (c) 2026, Institute for Protein Design, University of Washington. Optional third-party assets distributed alongside the model — the AlphaFold2 and OpenFold3/AlphaFold3 refolding images and the OpenFold3 payload — carry their own upstream terms; see the note under *Other repository files* and their provenance files. The training data has its own source terms, described on the [dataset card](https://huggingface.co/datasets/nvidia/RFD4-Proteina).
|
| 290 |
+
|
| 291 |
+
## Citation
|
| 292 |
+
|
| 293 |
+
When reporting results, cite the RFD4-Proteína project together with the exact model revision, code revision, seed, inputs and sampling configuration you used, plus any source datasets and evaluation tools.
|
| 294 |
+
|
| 295 |
+
```bibtex
|
| 296 |
+
@software{rfd4_proteina_2026,
|
| 297 |
+
title = {RFD4-Proteína: joint sequence and all-atom structure design},
|
| 298 |
+
author = {{Baker Lab, Institute for Protein Design, University of Washington} and {NVIDIA GenAIR}},
|
| 299 |
+
year = {2026},
|
| 300 |
+
url = {https://huggingface.co/nvidia/RFD4-Proteina},
|
| 301 |
+
note = {Model revision a06e2509c7033b010fc43614e9fd840303588502}
|
| 302 |
+
}
|
| 303 |
+
```
|
| 304 |
+
|
| 305 |
+
## Attribution and support
|
| 306 |
+
|
| 307 |
+
Developed by the Baker Lab at the Institute for Protein Design, University of Washington, and NVIDIA GenAIR.
|
| 308 |
+
|
| 309 |
+
[Code and issues](https://github.com/RosettaCommons/RFD4-Proteina/issues) · [Model history](https://huggingface.co/nvidia/RFD4-Proteina/commits/main) · [License](LICENSE)
|
adapter/adapter_config.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": null,
|
| 6 |
+
"base_model_name_or_path": null,
|
| 7 |
+
"bias": "none",
|
| 8 |
+
"corda_config": null,
|
| 9 |
+
"ensure_weight_tying": false,
|
| 10 |
+
"eva_config": null,
|
| 11 |
+
"exclude_modules": null,
|
| 12 |
+
"fan_in_fan_out": false,
|
| 13 |
+
"inference_mode": false,
|
| 14 |
+
"init_lora_weights": true,
|
| 15 |
+
"kasa_config": null,
|
| 16 |
+
"layer_replication": null,
|
| 17 |
+
"layers_pattern": null,
|
| 18 |
+
"layers_to_transform": null,
|
| 19 |
+
"loftq_config": {},
|
| 20 |
+
"lora_alpha": 64.0,
|
| 21 |
+
"lora_bias": false,
|
| 22 |
+
"lora_dropout": 0.0,
|
| 23 |
+
"lora_ga_config": null,
|
| 24 |
+
"megatron_config": null,
|
| 25 |
+
"megatron_core": "megatron.core",
|
| 26 |
+
"modules_to_save": null,
|
| 27 |
+
"monteclora_config": null,
|
| 28 |
+
"peft_type": "LORA",
|
| 29 |
+
"peft_version": "0.21.2",
|
| 30 |
+
"qalora_group_size": 16,
|
| 31 |
+
"r": 32,
|
| 32 |
+
"rank_pattern": {},
|
| 33 |
+
"revision": null,
|
| 34 |
+
"target_modules": [
|
| 35 |
+
"to_qkv",
|
| 36 |
+
"to_out_node"
|
| 37 |
+
],
|
| 38 |
+
"target_parameters": null,
|
| 39 |
+
"task_type": null,
|
| 40 |
+
"trainable_token_indices": null,
|
| 41 |
+
"use_bdlora": null,
|
| 42 |
+
"use_dora": false,
|
| 43 |
+
"use_qalora": false,
|
| 44 |
+
"use_rslora": false,
|
| 45 |
+
"velora_config": null
|
| 46 |
+
}
|
adapter/adapter_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b0762a73b543dfd88958267ba93a729fd91b31f2e18dba2d34c302260f50402
|
| 3 |
+
size 8264360
|
ae_checkpoints/rfd4_proteina_ae-ema.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:24536372f3fc9a5083b6b813baa80630d2287a5d61fbf2f126c9735802e1d21a
|
| 3 |
+
size 1029340246
|
ae_checkpoints/rfd4_proteina_ae-ema.ckpt.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
24536372f3fc9a5083b6b813baa80630d2287a5d61fbf2f126c9735802e1d21a rfd4_proteina_ae-ema.ckpt
|
alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sif
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:62b6531b87bafbbfb584ae2c341b9bab4591a733ffbc39abd2d6dd4b7ac214c6
|
| 3 |
+
size 16566288384
|
alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sif.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
62b6531b87bafbbfb584ae2c341b9bab4591a733ffbc39abd2d6dd4b7ac214c6 alphafold2-pallas-cuda12-amd64.sif
|
alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sqsh
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b037a78501b5d3c2d39ebdf9faecb3b4e0e7d7adc7bf4693f33699ba2b7ddc9
|
| 3 |
+
size 16566255616
|
alphafold2/runtime/alphafold2-pallas-cuda12-amd64.sqsh.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
8b037a78501b5d3c2d39ebdf9faecb3b4e0e7d7adc7bf4693f33699ba2b7ddc9 alphafold2-pallas-cuda12-amd64.sqsh
|
alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sif
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5160d179a0f11608e176f8883c8e9c644ab5c6a6b92bf05c2ca5b7a5ecb8d3f
|
| 3 |
+
size 6051827712
|
alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sif.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
b5160d179a0f11608e176f8883c8e9c644ab5c6a6b92bf05c2ca5b7a5ecb8d3f alphafold2-pallas-cuda13-amd64.sif
|
alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sqsh
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9982da8f06ee257366e31f395c0d006bb28ada4d22137e5561b5b0f908c9804a
|
| 3 |
+
size 12401594368
|
alphafold2/runtime/alphafold2-pallas-cuda13-amd64.sqsh.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
9982da8f06ee257366e31f395c0d006bb28ada4d22137e5561b5b0f908c9804a alphafold2-pallas-cuda13-amd64.sqsh
|
assets/design-gallery.png
ADDED
|
Git LFS Details
|
assets/examples/antibody_design.png
ADDED
|
Git LFS Details
|
assets/examples/enzyme_design.png
ADDED
|
Git LFS Details
|
assets/examples/protein_binder.png
ADDED
|
Git LFS Details
|
assets/wordmark-dark.webp
ADDED
|
assets/wordmark.webp
ADDED
|
base/config.json
ADDED
|
@@ -0,0 +1,263 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "rfd4_proteina",
|
| 3 |
+
"format_version": 1,
|
| 4 |
+
"weights_sha256": "10e90a76e3b09c94a700d2f67e89ed660af7f3018e3f2209b5d52f24d00fbd39",
|
| 5 |
+
"config": {
|
| 6 |
+
"nn": {
|
| 7 |
+
"name": "local_latents_transformer",
|
| 8 |
+
"output_parameterization": {
|
| 9 |
+
"bb_ca": "v",
|
| 10 |
+
"local_latents": "v"
|
| 11 |
+
},
|
| 12 |
+
"sequence_readout": {
|
| 13 |
+
"enabled": true,
|
| 14 |
+
"detach": true,
|
| 15 |
+
"num_classes": 147
|
| 16 |
+
},
|
| 17 |
+
"nlayers": 14,
|
| 18 |
+
"token_dim": 768,
|
| 19 |
+
"nheads": 12,
|
| 20 |
+
"parallel_mha_transition": false,
|
| 21 |
+
"token_type_emb_dim": 32,
|
| 22 |
+
"cath_emb_dim": 32,
|
| 23 |
+
"strict_feats": false,
|
| 24 |
+
"noise_align_chirality": true,
|
| 25 |
+
"noise_align_internal_coords": false,
|
| 26 |
+
"noise_align_chirality_translate": false,
|
| 27 |
+
"noise_align_internal_coords_translate": false,
|
| 28 |
+
"use_atom37_factory": true,
|
| 29 |
+
"feats_seq": [
|
| 30 |
+
"xt_bb_ca",
|
| 31 |
+
"xt_local_latents",
|
| 32 |
+
"token_type",
|
| 33 |
+
"sequence_probabilities",
|
| 34 |
+
"cond_a37coors_nm",
|
| 35 |
+
"cond_a37coors_internal_nm",
|
| 36 |
+
"rasa_binned",
|
| 37 |
+
"secondary_structure",
|
| 38 |
+
"hbond_modernized",
|
| 39 |
+
"pistack",
|
| 40 |
+
"pication",
|
| 41 |
+
"saltbridge",
|
| 42 |
+
"hydrophobic",
|
| 43 |
+
"metal_coord",
|
| 44 |
+
"na_ss_state",
|
| 45 |
+
"hotspot",
|
| 46 |
+
"atom_charge",
|
| 47 |
+
"atomic_element",
|
| 48 |
+
"coord_group_seq",
|
| 49 |
+
"chirality",
|
| 50 |
+
"chain_centroid_xyz"
|
| 51 |
+
],
|
| 52 |
+
"feats_cond_seq": [
|
| 53 |
+
"time_emb_bb_ca",
|
| 54 |
+
"time_emb_local_latents",
|
| 55 |
+
"cath_code",
|
| 56 |
+
"recycle_remaining"
|
| 57 |
+
],
|
| 58 |
+
"dim_cond": 256,
|
| 59 |
+
"idx_emb_dim": 256,
|
| 60 |
+
"t_emb_dim": 256,
|
| 61 |
+
"feats_pair_repr": [
|
| 62 |
+
"xt_bb_ca_pair_dists",
|
| 63 |
+
"bond_mask",
|
| 64 |
+
"bond_order",
|
| 65 |
+
"is_aromatic",
|
| 66 |
+
"index_group_pair",
|
| 67 |
+
"molecule_idx_pair",
|
| 68 |
+
"molecule_entity_pair",
|
| 69 |
+
"token_bond_path_length",
|
| 70 |
+
"canonical_polymer_ordering",
|
| 71 |
+
"na_ss_pair"
|
| 72 |
+
],
|
| 73 |
+
"feats_pair_cond": [
|
| 74 |
+
"time_emb_bb_ca",
|
| 75 |
+
"time_emb_local_latents"
|
| 76 |
+
],
|
| 77 |
+
"xt_pair_dist_dim": 30,
|
| 78 |
+
"xt_pair_dist_min": 0.1,
|
| 79 |
+
"xt_pair_dist_max": 3,
|
| 80 |
+
"seq_sep_dim": 64,
|
| 81 |
+
"pair_repr_dim": 384,
|
| 82 |
+
"update_pair_repr": true,
|
| 83 |
+
"update_pair_repr_every_n": 2,
|
| 84 |
+
"use_tri_mult": true,
|
| 85 |
+
"tri_mult_c": 256,
|
| 86 |
+
"tri_mult_bias": false,
|
| 87 |
+
"tri_mult_backend": "cuequivariance",
|
| 88 |
+
"recycle": {
|
| 89 |
+
"enabled": true,
|
| 90 |
+
"grad": "last",
|
| 91 |
+
"training_max_recycles": 9,
|
| 92 |
+
"condition_probability": 0.5
|
| 93 |
+
},
|
| 94 |
+
"checkpointing": "none",
|
| 95 |
+
"pair_write": "hadamard",
|
| 96 |
+
"use_qkln": true,
|
| 97 |
+
"attention_type": "torch",
|
| 98 |
+
"use_torch_compile": false,
|
| 99 |
+
"norm_type": "rmsnorm_torch_fused"
|
| 100 |
+
},
|
| 101 |
+
"product_flowmatcher": {
|
| 102 |
+
"bb_ca": {
|
| 103 |
+
"zero_com_noise": true,
|
| 104 |
+
"guidance_enabled": true,
|
| 105 |
+
"dim": 3,
|
| 106 |
+
"stochastic_centering_mode": "x_0",
|
| 107 |
+
"stochastic_centering_scale": 0.2,
|
| 108 |
+
"stochastic_centering_scope": "per_group",
|
| 109 |
+
"stochastic_centering_group_keys": [
|
| 110 |
+
"molecule_iid"
|
| 111 |
+
]
|
| 112 |
+
},
|
| 113 |
+
"local_latents": {
|
| 114 |
+
"zero_com_noise": false,
|
| 115 |
+
"guidance_enabled": true,
|
| 116 |
+
"dim": 32,
|
| 117 |
+
"stochastic_centering_mode": "none",
|
| 118 |
+
"stochastic_centering_scale": null
|
| 119 |
+
}
|
| 120 |
+
},
|
| 121 |
+
"loss": {
|
| 122 |
+
"signed_volume_weight": 2.5,
|
| 123 |
+
"sequence_prediction": {
|
| 124 |
+
"enabled": true,
|
| 125 |
+
"t_min": 0.0,
|
| 126 |
+
"t_max": 1.0
|
| 127 |
+
},
|
| 128 |
+
"t_distribution": {
|
| 129 |
+
"bb_ca": {
|
| 130 |
+
"name": "mix_unif_beta",
|
| 131 |
+
"p1": 1.9,
|
| 132 |
+
"p2": 1.0,
|
| 133 |
+
"p3": 0.02
|
| 134 |
+
},
|
| 135 |
+
"local_latents": {
|
| 136 |
+
"name": "mix_unif_beta",
|
| 137 |
+
"p1": 1.0,
|
| 138 |
+
"p2": 1.5,
|
| 139 |
+
"p3": 0.2
|
| 140 |
+
},
|
| 141 |
+
"shared_groups": null
|
| 142 |
+
}
|
| 143 |
+
},
|
| 144 |
+
"training": {
|
| 145 |
+
"exploration_k": 4,
|
| 146 |
+
"exploration_losses": [
|
| 147 |
+
"bb_ca",
|
| 148 |
+
"local_latents"
|
| 149 |
+
],
|
| 150 |
+
"n_recycle_max": 9,
|
| 151 |
+
"n_recycle_sampler": "poisson",
|
| 152 |
+
"n_recycle_lambda": 3.0,
|
| 153 |
+
"n_recycle_sync": true,
|
| 154 |
+
"n_recycle_val_loss": 3,
|
| 155 |
+
"motif_conditioning": false
|
| 156 |
+
},
|
| 157 |
+
"opt": {
|
| 158 |
+
"lr": 0.0001,
|
| 159 |
+
"muon_warmup_steps": 0,
|
| 160 |
+
"adam_warmup_steps": 10000,
|
| 161 |
+
"gradient_clip_val": 1.0,
|
| 162 |
+
"max_epochs": 10000000,
|
| 163 |
+
"accumulate_grad_batches": 1,
|
| 164 |
+
"dist_strategy": "ddp",
|
| 165 |
+
"steps_per_epoch": 2000,
|
| 166 |
+
"val_every_n_epochs": 10,
|
| 167 |
+
"val_start_epoch": 0,
|
| 168 |
+
"compute_val_loss": false,
|
| 169 |
+
"foldseek_timeout_minutes": 25,
|
| 170 |
+
"muon_exclude_param_patterns": [
|
| 171 |
+
"_factory\\.",
|
| 172 |
+
"^nn\\.(local_latents_linear|ca_linear|insertion_q_proj|insertion_k_proj)\\.",
|
| 173 |
+
"^nn\\.sequence_readout\\."
|
| 174 |
+
],
|
| 175 |
+
"muon_adjust_lr_fn": "match_rms_adamw",
|
| 176 |
+
"muon_grouped": true,
|
| 177 |
+
"num_sanity_val_steps": 2,
|
| 178 |
+
"coordinate_descent_tuning": false
|
| 179 |
+
},
|
| 180 |
+
"autoencoder_config": {
|
| 181 |
+
"nn_ae": {
|
| 182 |
+
"name": "transformer_ae_aa_rna_dna_glycan_ncaa",
|
| 183 |
+
"latent_z_dim": 32,
|
| 184 |
+
"allow_oxt": true,
|
| 185 |
+
"use_torch_compile": false,
|
| 186 |
+
"torch_compile_fullgraph": false,
|
| 187 |
+
"n_residue_types": 147,
|
| 188 |
+
"encoder": {
|
| 189 |
+
"nlayers": 12,
|
| 190 |
+
"token_dim": 768,
|
| 191 |
+
"nheads": 12,
|
| 192 |
+
"parallel_mha_transition": false,
|
| 193 |
+
"token_type_emb_dim": 32,
|
| 194 |
+
"strict_feats": false,
|
| 195 |
+
"feats_seq": [
|
| 196 |
+
"token_type",
|
| 197 |
+
"x1_a37coors_nm",
|
| 198 |
+
"x1_a37coors_nm_rel",
|
| 199 |
+
"chain_idx_seq",
|
| 200 |
+
"incident_bond_type_counts",
|
| 201 |
+
"atom37_element_charge_embedding"
|
| 202 |
+
],
|
| 203 |
+
"feats_cond_seq": null,
|
| 204 |
+
"dim_cond": 128,
|
| 205 |
+
"idx_emb_dim": 128,
|
| 206 |
+
"feats_pair_repr": [
|
| 207 |
+
"token_bond_path_length",
|
| 208 |
+
"canonical_polymer_ordering",
|
| 209 |
+
"molecule_idx_pair",
|
| 210 |
+
"x1_bb_pair_dists_nm"
|
| 211 |
+
],
|
| 212 |
+
"seq_sep_dim": 64,
|
| 213 |
+
"pair_repr_dim": 256,
|
| 214 |
+
"update_pair_repr": false,
|
| 215 |
+
"update_pair_repr_every_n": 3,
|
| 216 |
+
"use_tri_mult": false,
|
| 217 |
+
"use_attn_pair_bias": false,
|
| 218 |
+
"use_qkln": true,
|
| 219 |
+
"normalize_latent": false,
|
| 220 |
+
"latent_z_dim": 32,
|
| 221 |
+
"n_residue_types": 147,
|
| 222 |
+
"charge_emb_dim": 16
|
| 223 |
+
},
|
| 224 |
+
"decoder": {
|
| 225 |
+
"nlayers": 12,
|
| 226 |
+
"token_dim": 768,
|
| 227 |
+
"abs_coors": false,
|
| 228 |
+
"nheads": 12,
|
| 229 |
+
"token_type_emb_dim": 32,
|
| 230 |
+
"parallel_mha_transition": false,
|
| 231 |
+
"strict_feats": false,
|
| 232 |
+
"feats_seq": [
|
| 233 |
+
"ca_coors_nm",
|
| 234 |
+
"z_latent_seq",
|
| 235 |
+
"token_type",
|
| 236 |
+
"cond_residue_type"
|
| 237 |
+
],
|
| 238 |
+
"feats_cond_seq": null,
|
| 239 |
+
"dim_cond": 128,
|
| 240 |
+
"idx_emb_dim": 128,
|
| 241 |
+
"feats_pair_repr": [
|
| 242 |
+
"token_bond_path_length",
|
| 243 |
+
"canonical_polymer_ordering",
|
| 244 |
+
"molecule_idx_pair",
|
| 245 |
+
"ca_coors_nm_pair_dists"
|
| 246 |
+
],
|
| 247 |
+
"seq_sep_dim": 64,
|
| 248 |
+
"pair_repr_dim": 256,
|
| 249 |
+
"update_pair_repr": false,
|
| 250 |
+
"update_pair_repr_every_n": 3,
|
| 251 |
+
"use_tri_mult": false,
|
| 252 |
+
"use_attn_pair_bias": false,
|
| 253 |
+
"use_qkln": true,
|
| 254 |
+
"latent_z_dim": 32,
|
| 255 |
+
"n_residue_types": 147
|
| 256 |
+
}
|
| 257 |
+
}
|
| 258 |
+
},
|
| 259 |
+
"path_length_search_cutoffs": [
|
| 260 |
+
64
|
| 261 |
+
]
|
| 262 |
+
}
|
| 263 |
+
}
|
base/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10e90a76e3b09c94a700d2f67e89ed660af7f3018e3f2209b5d52f24d00fbd39
|
| 3 |
+
size 1742660596
|
benchmarks/METHODS.md
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Supplement methods
|
| 2 |
+
|
| 3 |
+
This release uses the frozen cohorts in `evidence/freeze.json`, not every run in the live catalog. It includes current Figure 2 methods and the additional RFD4 High conditioning arms. Antibody, HalluDesign, full-pipeline BindCraft and Genie3 search are excluded. The per-run evidence is under `models/<model>/provenance/`; exact native inputs and timing receipts take precedence over any current upstream default.
|
| 4 |
+
|
| 5 |
+
## Common accounting and timing
|
| 6 |
+
|
| 7 |
+
Each model × protocol × target has **200 original attempts**, including native failures and unmatched required motifs. We do not replace rejected originals or select outputs with the common AF3 evaluation. Protein-Hunter and RFD4 search have explicit intrinsic selection; those stages count toward their generation cost. Novo's two arms share original scaffolds and are not independent samples.
|
| 8 |
+
|
| 9 |
+
`generation_gpu_s` is the complete stored warm generation cost, including required sequence design, intrinsic folding/search, and native gaps inside that timed boundary. It excludes setup, loading, warmup and compilation where separately measured. It is not a diffusion-only timer. Batched costs are amortized across the batch's originals; batch shares are not individually measured latency. `timing_workers.csv` retains available setup and warm totals. `batch_size` is the measured actual batch, including tails—not a claim of maximal possible throughput. Unknown metadata remains blank. Exact-length/bucket compilation limitations are retained in executed receipts; the release does not assert that all models were equally optimized.
|
| 10 |
+
|
| 11 |
+
`af3_gpu_s` is separate common evaluation time. Default efficiency is the **arithmetic mean across targets of successes / (generation + AF3 GPU time)**, multiplied by 600 for per-10-GPU-minute units. Failed originals and their consumed time stay included. MPNN and native-fold columns are components already included in generation: **do not add them again**. A zero means the component did not run or used no GPU; blank means it was not separately measured. A missing AF3 cost excludes that configuration from an inclusive efficiency comparison, not its successes from rate comparisons.
|
| 12 |
+
|
| 13 |
+
Timing exceptions: ODesign retains the user-approved original-generation timing proxy, which includes its old MPNN step rather than separately adding corrected MPNN. Complexa fixed-small-molecule costs are registered batched remeasurements, not original batch-1 wall time. Historical RFD4 PPI all-atom uses the approved backbone timing proxy and has no measured AF3 cost. These are identified in `timing_basis`; they are not disguised as direct per-design measurements.
|
| 14 |
+
|
| 15 |
+
## Model recipes and adaptations
|
| 16 |
+
|
| 17 |
+
| Model | Executed generation and input/output considerations | Author reference |
|
| 18 |
+
|---|---|---|
|
| 19 |
+
| RFD4-Proteina | Joint sequence/structure generation with stored Low/Medium/High schedules; exact schedule/checkpoint comes from each native config. Search is a separate intrinsic workflow: AF2-Multimer for PPI and OF3 for the other benchmarks, with all search work timed. Preserve native motif correspondence; unmatched required motifs fail. Historical PPI-AA validation is explicitly separate from the newer BB campaign. | [RFD4 repository](https://github.com/RosettaCommons/RFD4-Proteina) |
|
| 20 |
+
| RFdiffusion3 | Native 200 steps, CFG off, step size 1.5; one LigandMPNN sequence at temperature 0.1, no downstream pipeline screening. DNA uses the DNA-finetuned checkpoint. Native JSON/derivatives preserve shared lengths and correspondence. Approved PPI OXT omission and native M0110 bond/charge/preprocessing losses are retained. Ligands can follow native reference conformers; no stereo repair. | [Foundry](https://github.com/RosettaCommons/foundry) |
|
| 21 |
+
| BoltzGen | Design: 500 steps/3 recycles; inverse folding: 200 steps/3 recycles, one final sequence. Original indices 0–99 use diverse, 100–199 adherence, preserving the native quarter settings across shards. No terminal folding/filtering/ranking. Atom hotspots become residue hotspots; PPI is all-atom. Seven heme cases use disclosed native bond-type projections; difficult chemistry uses source-derived archives, never author benchmark molecules. | [BoltzGen](https://github.com/HannesStark/boltzgen) |
|
| 22 |
+
| Proteina-Complexa | One 400-step joint draw, no additional MPNN/selection. PPI uses V1; AME V2/LoRA32; fixed small molecules use the corrected dedicated V1/LoRA32 ligand route. Native matcher is retained: unmatched motifs fail downstream. PPI and ligand inputs use native atom37/template features and residue-projected hotspots. No DNA or fully unfixed ligand route. | [Proteina-Complexa](https://github.com/NVIDIA-BioNeMo/Proteina-Complexa) |
|
| 23 |
+
| DISCO | Native diverse/max joint draw, 200 diffusion steps/four recycles. No extra sequence designer. SDF/JSON derives from shared CIF chemistry, never checked-in benchmark SDFs. Same-target/exact-length batches; all batch members retained. Native hydrogen/bond-order losses are documented, not repaired. | [DISCO](https://github.com/DISCO-design/DISCO) |
|
| 24 |
+
| Genie3 | One 100-step generative draw and one CA-MPNN sequence at 0.1. **No search or ColabFold.** Force the original hotspot, using target CA and source hotspot sidechain geometry, not extended contacts. Native CA-only outputs stay CA-only and are evaluated with binder CA RMSD. Some executed input receipts reference an earlier source release; exact executed files remain preserved, not silently rebound to a newer queue. | [Genie3](https://github.com/aqlaboratory/genie3) |
|
| 25 |
+
| BindCraft (single backbone) | One 75/45/5/15 optimization trajectory, then one soluble ProteinMPNN v_48_020 sequence at 0.1. Three continuation gates bypassed to retain bad trajectories. No post-MPNN fold/relax/filter/rank. Native independent batch 1; exact lengths warmed. Native interface-fixed residues retained. Output sidechains incompatible with the new sequence become unobserved rather than being grafted. | [BindCraft tutorial](https://github.com/martinpacesa/BindCraft/wiki/De-novo-binder-design-with-BindCraft) |
|
| 26 |
+
| BindCraft2 (single backbone) | One 50/25/45/5/15 trajectory, then one native soluble-MPNN redesign at 0.1. Continuation gates disabled; no independent validation or relaxation. Native independent batch 1, unchunked; resident AF2/MPNN pair, separately warmed signatures. Preserve exact native redesign constraints and report omitted incompatible sidechains. | [BindCraft2](https://github.com/PacesaLab/BindCraft2) |
|
| 27 |
+
| Protein-Hunter | Seven approved native refinement cycles, 100%-X initialization, 200 steps/3 recycles, MPNN 0.1. SolubleMPNN for PPI; LigandMPNN for ligands/DNA. Retain highest native ipTM cycle among those with ≤20% alanine. All intrinsic cycles are timed. Native rejection and writer/chemistry defects are not rescued. PPI guidance is residue hotspots and CA/CB/frame guidance, not hard sidechain enforcement. | [Protein-Hunter](https://github.com/yehlincho/Protein-Hunter) |
|
| 28 |
+
| ESMFold2 inversion | 150 inversion steps, two pinned models/four intrinsic hero critics. Critic 0 is predesignated; no common-AF3 ranking. Target sequence/residue hotspots, not target coordinates. Multichain extension preserves all target copies and native whole-complex ipTM, including target–target contributions. Intrinsic critics are not separate benchmark attempts. | [ESM binder-design tutorial](https://github.com/Biohub/esm/blob/main/cookbook/tutorials/binder_design.ipynb) |
|
| 29 |
+
| RFdiffusion | 50-step Complex_base draw with example CA/frame noise scales 0; one ProteinMPNN v_48_020 sequence at 0.1. Resident native batch 1. No Rosetta/refold/selection. Explicit contig/chain maps retain target copies and gaps. Preserve native backbone-only output and atom omissions. | [RFdiffusion binder design](https://github.com/RosettaCommons/RFdiffusion#binder-design) |
|
| 30 |
+
| RFdiffusion2 | AME RFD_140 checkpoint, 100 native flow steps, one LigandMPNN 0.1 sequence with native packing/context. Resident native batch 1. M0110 uses the explicitly approved disconnected ligand projection: preserve source atoms/coordinates but omit two bonds in the native derivative. This is not equivalent physical topology. Native guidepost matching/idealization is retained. | [RFdiffusion2](https://github.com/RosettaCommons/RFdiffusion2) |
|
| 31 |
+
| ODesign | Native scaffold generation followed by the corrected one-LigandMPNN0.1, source-mask-fixed sequence route. No terminal AF3 selection. Native AME receives disclosed extra observed motif context and is not a strict tip-only baseline. Fixed-ligand and fixed-DNA membership is labeled explicitly. Original-generation timing is a proxy, as described above. | [ODesign](https://github.com/OTeam-AI4S/ODesign) |
|
| 32 |
+
| AlphaProtein-novo | Native 1000-step joint codesign. Paired LigandMPNN arm uses one sequence at 0.1, checkpoint 030, atom context on, sidechain context off, no packing. CPU MPNN adds zero GPU seconds; shared scaffold generation time is inherited, not omitted. Native APN1 composite-ligand handling of M0110 changes interface interpretation and is disclosed. | [AlphaProtein-novo](https://github.com/google-deepmind/alphaprotein-novo) |
|
| 33 |
+
|
| 34 |
+
Native source/current workflow snapshots and immutable repository citations are listed in `references.csv`. They explain the implementation; executed receipts—not the newest README—identify what each cohort actually ran.
|
| 35 |
+
|
| 36 |
+
## Common AF3 scoring
|
| 37 |
+
|
| 38 |
+
The existing accelerated external-eval AF3 outputs are reused; this release launches no generation, folding or rescoring. Protein targets are templated from propagated coordinate-conditioning information using stored coordinates. DNA alignment includes the target DNA in both conditioning modes. Template sanity metrics are diagnostic, not substitutions for binder accuracy. Native biological-symmetry handling is retained in stored scores; do not derive alternative scores merely by relabeling chains.
|
| 39 |
+
|
| 40 |
+
| Benchmark | Current pass criterion |
|
| 41 |
+
|---|---|
|
| 42 |
+
| PPI, 24 targets | AF3 binder pTM ≥80 on the native 0–100 scale; minimum interface PAE ≤1.5 Å; target-backbone-aligned binder backbone RMSD ≤2.5 Å (CA-only for Genie3). Required target mapping/conditioning must be valid. Target mutations at a hotspot or within 5 Å of binder heavy atoms in design or refold fail; distal substitutions may pass. |
|
| 43 |
+
| Small molecule, 77 targets | AF3 minimum interface PAE ≤1.5 Å, chain-of-interest ipTM ≥80/100, protein-backbone RMSD ≤1.5 Å, and protein-backbone-aligned ligand RMSD ≤5 Å. This is `sm_binder`, not the looser historical `disco` criterion. |
|
| 44 |
+
| Original DISCO, same 77 targets | Protein-backbone RMSD ≤2 Å and maximum ligand-centroid displacement ≤2 Å after protein-backbone alignment. No AF3 PAE/ipTM cutoff. Stored verdicts and independently computed passing-cluster labels are retained separately from `sm_binder`. The criteria are not mathematically nested because the ligand-position tests differ. |
|
| 45 |
+
| Atomic Motif Enzyme (AME), 37 targets | Protein-backbone RMSD ≤2 Å, required motif all-atom RMSD ≤1.5 Å, and AF3 minimum interface PAE **<1.5 Å**. Motif RMSD aligns/scores mapped conditioned protein atoms against stored input references, respecting coordinate groups. Native-unmatched/sequence-failed motifs fail; missing metrics are not assigned zero. Clashes are diagnostic only in this current policy. |
|
| 46 |
+
| DNA, 3 targets | AF3 minimum interface PAE ≤1.5 Å and DNA-target-aligned binder backbone RMSD ≤8 Å. Alignment uses all target DNA atoms in both fixed/unfixed modes. |
|
| 47 |
+
|
| 48 |
+
Ordinary rate = passing originals / 200. Clustered rate = distinct clusters among passing originals / 200. Clusters are computed **within each model/protocol/target**, using generated binder/protein structures and Foldseek TM threshold 0.5, alignment type 1, minimum sequence identity 0; they are not pooled between targets or models. Dark bar plus light extension is the ordinary rate; dark alone is clustered. These are algorithmic structural clusters, not experimental successes or globally distinct proteins.
|
| 49 |
+
|
| 50 |
+
## Supplement selection and conditioning comparisons
|
| 51 |
+
|
| 52 |
+
Per-target panels include RFD4 High, the best other generative model, and best non-generative model by mean clustered rate (ordinary rate breaks ties). The requested AME panel instead shows four: RFD4 High, AlphaProtein-novo Codesign, Proteina-Complexa and RFdiffusion3. Both small-molecule criteria get separate panels and comparator selection. Full selection and all model values remain in CSVs.
|
| 53 |
+
|
| 54 |
+
The High conditioning panels use all 200 originals per target and do not assume the desired direction. **PPI-AA and PPI-BB are different campaigns**; other settings/checkpoints can confound conditioning. AME compares *partially* to fully fixed ligand, not fully unfixed. Three NAD small-molecule cases use different input conformers. Therefore these plots show observed differences, not isolated causal effects of removing constraints.
|
| 55 |
+
|
| 56 |
+
Historical originals with unavailable separate native/raw/AF3 artifacts remain represented. `artifacts-audited.csv` and `input_coverage.csv` distinguish available files from unresolved or intentionally retired evidence; a missing file is not a new failed design.
|
benchmarks/provenance.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"source_release": "supplement-20261009",
|
| 3 |
+
"paper_commit": "8614542c149875b13ed3fed5d8d6ec739f2daff4",
|
| 4 |
+
"source_path": "paper/figures/figure 2/benchmark_packet/supplement",
|
| 5 |
+
"metric": "mean_over_targets(3600 * passing_structure_clusters / (generation_gpu_s + af3_gpu_s))",
|
| 6 |
+
"summary_rows_recomputed": 52,
|
| 7 |
+
"target_criterion_rows": 2488,
|
| 8 |
+
"sha256": {
|
| 9 |
+
"summary.csv": "5c6d525d988d0df7fecb74396f86401c1ddb4c31fa83f00b357ba77d855d8a10",
|
| 10 |
+
"source_data.csv": "5f71454272467ea86a5ef0f4ad8a7c781ca31eff8d57497e83c87aa8820cc34b"
|
| 11 |
+
},
|
| 12 |
+
"scope": "Frozen project benchmark runs; not a new evaluation of every model Hub revision."
|
| 13 |
+
}
|
benchmarks/source_data.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
benchmarks/summary.csv
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
panel,model,protocol,dataset,label,criterion,pass_percent,clustered_percent,successes_per_10_gpu_min,unique_per_10_gpu_min,targets,attempts,timing_note
|
| 2 |
+
ame,alphaprotein_novo,one_shot,mcsa_fixed,AlphaProtein-novo (Codesign),ame,7.986486486486487,5.675675675675675,6.128165410759252,4.336287958764792,37,7400,per_attempt
|
| 3 |
+
ame,alphaprotein_novo_ligandmpnn,one_shot,mcsa_fixed,AlphaProtein-novo (LigandMPNN),ame,7.418918918918919,5.418918918918919,5.699414945101433,4.143206492049179,37,7400,inherited_parent;inherited_parent_generation_CPU_MPNN
|
| 4 |
+
ame,odesign,one_shot,mcsa_fixed,ODesign,ame,0.9054054054054054,0.2702702702702703,0.6290490961497859,0.1834221580606205,37,7400,per_attempt;original_generation_timing_proxy_v1
|
| 5 |
+
ame,pipelines,one_shot,mcsa_fixed,RFdiffusion3,ame,2.864864864864865,2.27027027027027,3.6288471843745116,2.873393749308303,37,7400,per_attempt
|
| 6 |
+
ame,proteina_complexa,one_shot,mcsa_fixed,Proteina-Complexa,ame,3.283783783783784,2.324324324324324,6.040088839302828,4.217084980960173,37,7400,per_attempt
|
| 7 |
+
ame,rfd4_proteina_high,one_shot,mcsa_fixed,RFD4-Proteina (high thinking),ame,13.445945945945946,8.905405405405405,23.248249750649716,15.197139241828658,37,7400,per_attempt
|
| 8 |
+
ame,rfd4_proteina_high_search,native_pipeline,mcsa_fixed,RFD4-Proteina (search),ame,17.81081081081081,12.432432432432432,2.1139479255711926,1.4702629167109056,37,7400,per_attempt
|
| 9 |
+
ame,rfd4_proteina_low,one_shot,mcsa_fixed,RFD4-Proteina (low thinking),ame,10.62162162162162,7.108108108108108,24.61903561723674,16.21640377305806,37,7400,per_attempt
|
| 10 |
+
ame,rfd4_proteina_medium,one_shot,mcsa_fixed,RFD4-Proteina (medium thinking),ame,12.37837837837838,8.216216216216216,26.60857770347569,17.39937846539931,37,7400,per_attempt
|
| 11 |
+
ame,rfdiffusion2,one_shot,mcsa_fixed,RFdiffusion2,ame,0.8378378378378378,0.7837837837837838,0.054484078041217035,0.05061505532429724,37,7400,per_attempt
|
| 12 |
+
dna,boltzgen,one_shot,dna_fixed,BoltzGen (fixed),dna,0.8333333333333334,0.8333333333333334,1.5767861157347076,1.5767861157347076,3,600,per_attempt
|
| 13 |
+
dna,odesign,one_shot,dna_fixed,ODesign (fixed),dna,0.0,0.0,0.0,0.0,3,600,per_attempt
|
| 14 |
+
dna,pipelines,one_shot,dna_fixed,RFdiffusion3 (fixed),dna,0.1666666666666666,0.1666666666666666,0.12549701345123473,0.12549701345123473,3,600,per_attempt
|
| 15 |
+
dna,protein_hunter,native_pipeline,dna_unfixed,Protein-Hunter,dna,0.3333333333333333,0.3333333333333333,0.0339968718491103,0.0339968718491103,3,600,per_attempt
|
| 16 |
+
dna,rfd4_proteina_high,one_shot,dna_unfixed,RFD4-Proteina (high thinking),dna,2.333333333333333,2.1666666666666665,7.533663081020337,6.995141754963048,3,600,per_attempt
|
| 17 |
+
dna,rfd4_proteina_high_search,native_pipeline,dna_unfixed,RFD4-Proteina (search),dna,3.8333333333333335,3.5,0.6977291779821485,0.637527788137053,3,600,per_attempt
|
| 18 |
+
dna,rfd4_proteina_low,one_shot,dna_unfixed,RFD4-Proteina (low thinking),dna,1.3333333333333333,1.3333333333333333,4.501005002182411,4.501005002182411,3,600,per_attempt
|
| 19 |
+
dna,rfd4_proteina_medium,one_shot,dna_unfixed,RFD4-Proteina (medium thinking),dna,2.0,1.6666666666666667,6.57941016603549,5.484803149401414,3,600,per_attempt
|
| 20 |
+
ppi,bindcraft,one_shot,ppi_all_atom,BindCraft (single backbone),ppi,9.9375,5.854166666666667,0.6858558193759934,0.38947020233767843,24,4800,per_attempt
|
| 21 |
+
ppi,bindcraft2,one_shot,ppi_all_atom,BindCraft2 (single backbone),ppi,7.229166666666667,4.8125,0.5281467210933602,0.3331281596040387,24,4800,per_attempt
|
| 22 |
+
ppi,boltzgen,one_shot,ppi_all_atom,BoltzGen,ppi,4.8125,1.2916666666666667,5.406141349312153,1.3833348139931438,24,4800,per_attempt
|
| 23 |
+
ppi,esmfold2,one_shot,ppi_backbone,ESMFold2,ppi,8.916666666666666,2.458333333333333,0.7509894473902138,0.19387923161633483,24,4800,per_attempt
|
| 24 |
+
ppi,genie3,one_shot,ppi_all_atom,Genie3,ppi,0.6875,0.5208333333333334,0.5547110110121317,0.42859016992113674,24,4800,per_attempt
|
| 25 |
+
ppi,odesign,one_shot,ppi_all_atom,ODesign,ppi,1.4583333333333333,1.2916666666666667,0.7635954268204376,0.668254791590555,24,4800,per_attempt;original_generation_timing_proxy_v1
|
| 26 |
+
ppi,pipelines,one_shot,ppi_all_atom,RFdiffusion3,ppi,3.375,1.7916666666666667,2.203438218078908,1.1140170381975611,24,4800,per_attempt
|
| 27 |
+
ppi,protein_hunter,native_pipeline,ppi_all_atom,Protein-Hunter,ppi,6.916666666666667,4.145833333333333,0.5573716351941015,0.33134699754642055,24,4800,per_attempt
|
| 28 |
+
ppi,proteina_complexa,one_shot,ppi_all_atom,Proteina-Complexa,ppi,4.5,1.8541666666666667,5.446523655582822,2.207096244584333,24,4800,per_attempt
|
| 29 |
+
ppi,rfd4-search-multimer,native_pipeline,ppi_backbone,RFD4-Proteina (search),ppi,24.45833333333333,11.916666666666666,1.7550251894828512,0.8465949585039426,24,4800,per_attempt
|
| 30 |
+
ppi,rfd4_proteina_high,one_shot,ppi_backbone,RFD4-Proteina (high thinking),ppi,19.14583333333333,9.770833333333334,24.50772631322702,11.933874547020181,24,4800,batch_amortized;batched_generation_remeasurement_v1
|
| 31 |
+
ppi,rfd4_proteina_low,one_shot,ppi_backbone,RFD4-Proteina (low thinking),ppi,14.104166666666666,7.041666666666667,23.666769202318935,11.560507426898935,24,4800,batch_amortized;batched_generation_remeasurement_v1
|
| 32 |
+
ppi,rfd4_proteina_medium,one_shot,ppi_backbone,RFD4-Proteina (medium thinking),ppi,16.395833333333332,8.354166666666666,25.812263933585978,12.73345240158931,24,4800,batch_amortized;batched_generation_remeasurement_v1
|
| 33 |
+
ppi,rfdiffusion,one_shot,ppi_all_atom,RFdiffusion,ppi,2.6041666666666665,0.75,0.5691347855847814,0.15660421612078013,24,4800,per_attempt
|
| 34 |
+
small_molecule,boltzgen,one_shot,studio_figure_2c_unfixed,BoltzGen,sm_binder,4.045454545454546,1.318181818181818,8.145222361450198,2.579619122276541,77,15400,per_attempt
|
| 35 |
+
small_molecule,disco,one_shot,studio_figure_2c_unfixed,DISCO,sm_binder,2.551948051948052,1.5714285714285714,0.20401419810891713,0.12533260451848793,77,15400,per_attempt
|
| 36 |
+
small_molecule,odesign,one_shot,studio_figure_2c_fixed,ODesign (fixed),sm_binder,0.0974025974025974,0.0519480519480519,0.06812084076412611,0.0366360091773046,77,15400,per_attempt
|
| 37 |
+
small_molecule,pipelines,one_shot,studio_figure_2c_unfixed,RFdiffusion3,sm_binder,3.512987012987013,2.3376623376623376,3.2330633144392786,2.1472317944116814,77,15400,per_attempt
|
| 38 |
+
small_molecule,protein_hunter,native_pipeline,studio_figure_2c_unfixed,Protein-Hunter,sm_binder,0.7857142857142857,0.6688311688311688,0.08222397901207588,0.06994325849054965,77,15400,per_attempt
|
| 39 |
+
small_molecule,proteina_complexa,one_shot,studio_figure_2c_fixed,Proteina-Complexa (fixed),sm_binder,0.1493506493506493,0.1428571428571428,0.3550431740456386,0.3394022404125218,77,15400,batch_amortized;batched_generation_remeasurement_v1
|
| 40 |
+
small_molecule,rfd4_proteina_high,one_shot,studio_figure_2c_unfixed,RFD4-Proteina (high thinking),sm_binder,27.818181818181817,8.74025974025974,59.80580730283336,18.708070088438323,77,15400,per_attempt
|
| 41 |
+
small_molecule,rfd4_proteina_high_search,native_pipeline,studio_figure_2c_unfixed,RFD4-Proteina (search),sm_binder,40.29220779220779,12.863636363636363,6.4266313688585415,2.042660497770513,77,15400,per_attempt
|
| 42 |
+
small_molecule,rfd4_proteina_low,one_shot,studio_figure_2c_unfixed,RFD4-Proteina (low thinking),sm_binder,20.92207792207792,6.37012987012987,58.905492101587086,17.847324331293226,77,15400,per_attempt
|
| 43 |
+
small_molecule,rfd4_proteina_medium,one_shot,studio_figure_2c_unfixed,RFD4-Proteina (medium thinking),sm_binder,25.18831168831169,7.577922077922078,66.53565733313984,19.911485476039143,77,15400,per_attempt
|
| 44 |
+
small_molecule_original,boltzgen,one_shot,studio_figure_2c_unfixed,BoltzGen,disco,27.675324675324678,6.097402597402597,53.78792519211063,11.761316579129796,77,15400,per_attempt
|
| 45 |
+
small_molecule_original,disco,one_shot,studio_figure_2c_unfixed,DISCO,disco,31.32467532467533,10.474025974025974,2.4539028999553714,0.8175381293508085,77,15400,per_attempt
|
| 46 |
+
small_molecule_original,odesign,one_shot,studio_figure_2c_fixed,ODesign (fixed),disco,1.5909090909090908,1.025974025974026,1.1154226419662379,0.7190330690062327,77,15400,per_attempt
|
| 47 |
+
small_molecule_original,pipelines,one_shot,studio_figure_2c_unfixed,RFdiffusion3,disco,26.13636363636364,12.844155844155845,23.779749576062088,11.679276621280051,77,15400,per_attempt
|
| 48 |
+
small_molecule_original,protein_hunter,native_pipeline,studio_figure_2c_unfixed,Protein-Hunter,disco,3.961038961038961,2.415584415584416,0.41375915688116066,0.2519214097355728,77,15400,per_attempt
|
| 49 |
+
small_molecule_original,proteina_complexa,one_shot,studio_figure_2c_fixed,Proteina-Complexa (fixed),disco,14.818181818181818,12.175324675324676,32.58468485951315,26.699142756356586,77,15400,batch_amortized;batched_generation_remeasurement_v1
|
| 50 |
+
small_molecule_original,rfd4_proteina_high,one_shot,studio_figure_2c_unfixed,RFD4-Proteina (high thinking),disco,63.25974025974026,13.915584415584416,136.2596016660901,29.843032429691466,77,15400,per_attempt
|
| 51 |
+
small_molecule_original,rfd4_proteina_high_search,native_pipeline,studio_figure_2c_unfixed,RFD4-Proteina (search),disco,73.35064935064935,18.11688311688312,11.821053066769755,2.892460870734589,77,15400,per_attempt
|
| 52 |
+
small_molecule_original,rfd4_proteina_low,one_shot,studio_figure_2c_unfixed,RFD4-Proteina (low thinking),disco,57.18181818181818,11.915584415584416,160.78300033866194,33.311256689662514,77,15400,per_attempt
|
| 53 |
+
small_molecule_original,rfd4_proteina_medium,one_shot,studio_figure_2c_unfixed,RFD4-Proteina (medium thinking),disco,61.0974025974026,12.474025974025974,161.40828898514403,32.75934584692291,77,15400,per_attempt
|
config.json
ADDED
|
@@ -0,0 +1,263 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "rfd4_proteina",
|
| 3 |
+
"format_version": 1,
|
| 4 |
+
"weights_sha256": "5b549dd61adb2f6c6b5cb95bda13f40f5e31556b4e6c755edc0ca6f011c68fb0",
|
| 5 |
+
"config": {
|
| 6 |
+
"nn": {
|
| 7 |
+
"name": "local_latents_transformer",
|
| 8 |
+
"output_parameterization": {
|
| 9 |
+
"bb_ca": "v",
|
| 10 |
+
"local_latents": "v"
|
| 11 |
+
},
|
| 12 |
+
"sequence_readout": {
|
| 13 |
+
"enabled": true,
|
| 14 |
+
"detach": true,
|
| 15 |
+
"num_classes": 147
|
| 16 |
+
},
|
| 17 |
+
"nlayers": 14,
|
| 18 |
+
"token_dim": 768,
|
| 19 |
+
"nheads": 12,
|
| 20 |
+
"parallel_mha_transition": false,
|
| 21 |
+
"token_type_emb_dim": 32,
|
| 22 |
+
"cath_emb_dim": 32,
|
| 23 |
+
"strict_feats": false,
|
| 24 |
+
"noise_align_chirality": true,
|
| 25 |
+
"noise_align_internal_coords": false,
|
| 26 |
+
"noise_align_chirality_translate": false,
|
| 27 |
+
"noise_align_internal_coords_translate": false,
|
| 28 |
+
"use_atom37_factory": true,
|
| 29 |
+
"feats_seq": [
|
| 30 |
+
"xt_bb_ca",
|
| 31 |
+
"xt_local_latents",
|
| 32 |
+
"token_type",
|
| 33 |
+
"sequence_probabilities",
|
| 34 |
+
"cond_a37coors_nm",
|
| 35 |
+
"cond_a37coors_internal_nm",
|
| 36 |
+
"rasa_binned",
|
| 37 |
+
"secondary_structure",
|
| 38 |
+
"hbond_modernized",
|
| 39 |
+
"pistack",
|
| 40 |
+
"pication",
|
| 41 |
+
"saltbridge",
|
| 42 |
+
"hydrophobic",
|
| 43 |
+
"metal_coord",
|
| 44 |
+
"na_ss_state",
|
| 45 |
+
"hotspot",
|
| 46 |
+
"atom_charge",
|
| 47 |
+
"atomic_element",
|
| 48 |
+
"coord_group_seq",
|
| 49 |
+
"chirality",
|
| 50 |
+
"chain_centroid_xyz"
|
| 51 |
+
],
|
| 52 |
+
"feats_cond_seq": [
|
| 53 |
+
"time_emb_bb_ca",
|
| 54 |
+
"time_emb_local_latents",
|
| 55 |
+
"cath_code",
|
| 56 |
+
"recycle_remaining"
|
| 57 |
+
],
|
| 58 |
+
"dim_cond": 256,
|
| 59 |
+
"idx_emb_dim": 256,
|
| 60 |
+
"t_emb_dim": 256,
|
| 61 |
+
"feats_pair_repr": [
|
| 62 |
+
"xt_bb_ca_pair_dists",
|
| 63 |
+
"bond_mask",
|
| 64 |
+
"bond_order",
|
| 65 |
+
"is_aromatic",
|
| 66 |
+
"index_group_pair",
|
| 67 |
+
"molecule_idx_pair",
|
| 68 |
+
"molecule_entity_pair",
|
| 69 |
+
"token_bond_path_length",
|
| 70 |
+
"canonical_polymer_ordering",
|
| 71 |
+
"na_ss_pair"
|
| 72 |
+
],
|
| 73 |
+
"feats_pair_cond": [
|
| 74 |
+
"time_emb_bb_ca",
|
| 75 |
+
"time_emb_local_latents"
|
| 76 |
+
],
|
| 77 |
+
"xt_pair_dist_dim": 30,
|
| 78 |
+
"xt_pair_dist_min": 0.1,
|
| 79 |
+
"xt_pair_dist_max": 3,
|
| 80 |
+
"seq_sep_dim": 64,
|
| 81 |
+
"pair_repr_dim": 384,
|
| 82 |
+
"update_pair_repr": true,
|
| 83 |
+
"update_pair_repr_every_n": 2,
|
| 84 |
+
"use_tri_mult": true,
|
| 85 |
+
"tri_mult_c": 256,
|
| 86 |
+
"tri_mult_bias": false,
|
| 87 |
+
"tri_mult_backend": "cuequivariance",
|
| 88 |
+
"recycle": {
|
| 89 |
+
"enabled": true,
|
| 90 |
+
"grad": "last",
|
| 91 |
+
"training_max_recycles": 9,
|
| 92 |
+
"condition_probability": 0.5
|
| 93 |
+
},
|
| 94 |
+
"checkpointing": "none",
|
| 95 |
+
"pair_write": "hadamard",
|
| 96 |
+
"use_qkln": true,
|
| 97 |
+
"attention_type": "torch",
|
| 98 |
+
"use_torch_compile": false,
|
| 99 |
+
"norm_type": "rmsnorm_torch_fused"
|
| 100 |
+
},
|
| 101 |
+
"product_flowmatcher": {
|
| 102 |
+
"bb_ca": {
|
| 103 |
+
"zero_com_noise": true,
|
| 104 |
+
"guidance_enabled": true,
|
| 105 |
+
"dim": 3,
|
| 106 |
+
"stochastic_centering_mode": "x_0",
|
| 107 |
+
"stochastic_centering_scale": 0.2,
|
| 108 |
+
"stochastic_centering_scope": "per_group",
|
| 109 |
+
"stochastic_centering_group_keys": [
|
| 110 |
+
"molecule_iid"
|
| 111 |
+
]
|
| 112 |
+
},
|
| 113 |
+
"local_latents": {
|
| 114 |
+
"zero_com_noise": false,
|
| 115 |
+
"guidance_enabled": true,
|
| 116 |
+
"dim": 32,
|
| 117 |
+
"stochastic_centering_mode": "none",
|
| 118 |
+
"stochastic_centering_scale": null
|
| 119 |
+
}
|
| 120 |
+
},
|
| 121 |
+
"loss": {
|
| 122 |
+
"signed_volume_weight": 2.5,
|
| 123 |
+
"sequence_prediction": {
|
| 124 |
+
"enabled": true,
|
| 125 |
+
"t_min": 0.0,
|
| 126 |
+
"t_max": 1.0
|
| 127 |
+
},
|
| 128 |
+
"t_distribution": {
|
| 129 |
+
"bb_ca": {
|
| 130 |
+
"name": "mix_unif_beta",
|
| 131 |
+
"p1": 1.9,
|
| 132 |
+
"p2": 1.0,
|
| 133 |
+
"p3": 0.02
|
| 134 |
+
},
|
| 135 |
+
"local_latents": {
|
| 136 |
+
"name": "mix_unif_beta",
|
| 137 |
+
"p1": 1.0,
|
| 138 |
+
"p2": 1.5,
|
| 139 |
+
"p3": 0.2
|
| 140 |
+
},
|
| 141 |
+
"shared_groups": null
|
| 142 |
+
}
|
| 143 |
+
},
|
| 144 |
+
"training": {
|
| 145 |
+
"exploration_k": 4,
|
| 146 |
+
"exploration_losses": [
|
| 147 |
+
"bb_ca",
|
| 148 |
+
"local_latents"
|
| 149 |
+
],
|
| 150 |
+
"n_recycle_max": 9,
|
| 151 |
+
"n_recycle_sampler": "poisson",
|
| 152 |
+
"n_recycle_lambda": 3.0,
|
| 153 |
+
"n_recycle_sync": true,
|
| 154 |
+
"n_recycle_val_loss": 3,
|
| 155 |
+
"motif_conditioning": false
|
| 156 |
+
},
|
| 157 |
+
"opt": {
|
| 158 |
+
"lr": 5e-05,
|
| 159 |
+
"muon_warmup_steps": 0,
|
| 160 |
+
"adam_warmup_steps": 10000,
|
| 161 |
+
"gradient_clip_val": 1.0,
|
| 162 |
+
"max_epochs": 10000000,
|
| 163 |
+
"accumulate_grad_batches": 2,
|
| 164 |
+
"dist_strategy": "ddp",
|
| 165 |
+
"steps_per_epoch": 1000,
|
| 166 |
+
"val_every_n_epochs": 5,
|
| 167 |
+
"val_start_epoch": 0,
|
| 168 |
+
"compute_val_loss": false,
|
| 169 |
+
"foldseek_timeout_minutes": 25,
|
| 170 |
+
"muon_exclude_param_patterns": [
|
| 171 |
+
"_factory\\.",
|
| 172 |
+
"^nn\\.(local_latents_linear|ca_linear|insertion_q_proj|insertion_k_proj)\\.",
|
| 173 |
+
"^nn\\.sequence_readout\\."
|
| 174 |
+
],
|
| 175 |
+
"muon_adjust_lr_fn": "match_rms_adamw",
|
| 176 |
+
"muon_grouped": true,
|
| 177 |
+
"num_sanity_val_steps": 2,
|
| 178 |
+
"coordinate_descent_tuning": false
|
| 179 |
+
},
|
| 180 |
+
"autoencoder_config": {
|
| 181 |
+
"nn_ae": {
|
| 182 |
+
"name": "transformer_ae_aa_rna_dna_glycan_ncaa",
|
| 183 |
+
"latent_z_dim": 32,
|
| 184 |
+
"allow_oxt": true,
|
| 185 |
+
"use_torch_compile": false,
|
| 186 |
+
"torch_compile_fullgraph": false,
|
| 187 |
+
"n_residue_types": 147,
|
| 188 |
+
"encoder": {
|
| 189 |
+
"nlayers": 12,
|
| 190 |
+
"token_dim": 768,
|
| 191 |
+
"nheads": 12,
|
| 192 |
+
"parallel_mha_transition": false,
|
| 193 |
+
"token_type_emb_dim": 32,
|
| 194 |
+
"strict_feats": false,
|
| 195 |
+
"feats_seq": [
|
| 196 |
+
"token_type",
|
| 197 |
+
"x1_a37coors_nm",
|
| 198 |
+
"x1_a37coors_nm_rel",
|
| 199 |
+
"chain_idx_seq",
|
| 200 |
+
"incident_bond_type_counts",
|
| 201 |
+
"atom37_element_charge_embedding"
|
| 202 |
+
],
|
| 203 |
+
"feats_cond_seq": null,
|
| 204 |
+
"dim_cond": 128,
|
| 205 |
+
"idx_emb_dim": 128,
|
| 206 |
+
"feats_pair_repr": [
|
| 207 |
+
"token_bond_path_length",
|
| 208 |
+
"canonical_polymer_ordering",
|
| 209 |
+
"molecule_idx_pair",
|
| 210 |
+
"x1_bb_pair_dists_nm"
|
| 211 |
+
],
|
| 212 |
+
"seq_sep_dim": 64,
|
| 213 |
+
"pair_repr_dim": 256,
|
| 214 |
+
"update_pair_repr": false,
|
| 215 |
+
"update_pair_repr_every_n": 3,
|
| 216 |
+
"use_tri_mult": false,
|
| 217 |
+
"use_attn_pair_bias": false,
|
| 218 |
+
"use_qkln": true,
|
| 219 |
+
"normalize_latent": false,
|
| 220 |
+
"latent_z_dim": 32,
|
| 221 |
+
"n_residue_types": 147,
|
| 222 |
+
"charge_emb_dim": 16
|
| 223 |
+
},
|
| 224 |
+
"decoder": {
|
| 225 |
+
"nlayers": 12,
|
| 226 |
+
"token_dim": 768,
|
| 227 |
+
"abs_coors": false,
|
| 228 |
+
"nheads": 12,
|
| 229 |
+
"token_type_emb_dim": 32,
|
| 230 |
+
"parallel_mha_transition": false,
|
| 231 |
+
"strict_feats": false,
|
| 232 |
+
"feats_seq": [
|
| 233 |
+
"ca_coors_nm",
|
| 234 |
+
"z_latent_seq",
|
| 235 |
+
"token_type",
|
| 236 |
+
"cond_residue_type"
|
| 237 |
+
],
|
| 238 |
+
"feats_cond_seq": null,
|
| 239 |
+
"dim_cond": 128,
|
| 240 |
+
"idx_emb_dim": 128,
|
| 241 |
+
"feats_pair_repr": [
|
| 242 |
+
"token_bond_path_length",
|
| 243 |
+
"canonical_polymer_ordering",
|
| 244 |
+
"molecule_idx_pair",
|
| 245 |
+
"ca_coors_nm_pair_dists"
|
| 246 |
+
],
|
| 247 |
+
"seq_sep_dim": 64,
|
| 248 |
+
"pair_repr_dim": 256,
|
| 249 |
+
"update_pair_repr": false,
|
| 250 |
+
"update_pair_repr_every_n": 3,
|
| 251 |
+
"use_tri_mult": false,
|
| 252 |
+
"use_attn_pair_bias": false,
|
| 253 |
+
"use_qkln": true,
|
| 254 |
+
"latent_z_dim": 32,
|
| 255 |
+
"n_residue_types": 147
|
| 256 |
+
}
|
| 257 |
+
}
|
| 258 |
+
},
|
| 259 |
+
"path_length_search_cutoffs": [
|
| 260 |
+
64
|
| 261 |
+
]
|
| 262 |
+
}
|
| 263 |
+
}
|
examples/antibody_design/design.cif
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
examples/antibody_design/input.cif
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
examples/enzyme_design/design.cif
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
examples/enzyme_design/input.cif
ADDED
|
@@ -0,0 +1,177 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data_M0630_1j79.cif
|
| 2 |
+
#
|
| 3 |
+
_entry.id M0630_1j79.cif
|
| 4 |
+
_entry.author rib7
|
| 5 |
+
_entry.date 2026-08-28
|
| 6 |
+
_entry.time 20:38:54
|
| 7 |
+
#
|
| 8 |
+
loop_
|
| 9 |
+
_entity.id
|
| 10 |
+
_entity.type
|
| 11 |
+
_entity.pdbx_description
|
| 12 |
+
0 polymer .
|
| 13 |
+
1 non-polymer .
|
| 14 |
+
2 non-polymer .
|
| 15 |
+
#
|
| 16 |
+
_entity_poly.entity_id 0
|
| 17 |
+
_entity_poly.type polypeptide(L)
|
| 18 |
+
_entity_poly.nstd_linkage no
|
| 19 |
+
_entity_poly.nstd_monomer yes
|
| 20 |
+
_entity_poly.pdbx_seq_one_letter_code XD
|
| 21 |
+
_entity_poly.pdbx_seq_one_letter_code_can XD
|
| 22 |
+
_entity_poly.pdbx_strand_id A
|
| 23 |
+
_entity_poly.pdbx_target_identifier ?
|
| 24 |
+
#
|
| 25 |
+
loop_
|
| 26 |
+
_entity_poly_seq.entity_id
|
| 27 |
+
_entity_poly_seq.hetero
|
| 28 |
+
_entity_poly_seq.mon_id
|
| 29 |
+
_entity_poly_seq.num
|
| 30 |
+
0 n 180-180 0
|
| 31 |
+
0 n ASP 1
|
| 32 |
+
#
|
| 33 |
+
loop_
|
| 34 |
+
_chem_comp.id
|
| 35 |
+
_chem_comp.name
|
| 36 |
+
_chem_comp.type
|
| 37 |
+
_chem_comp.formula
|
| 38 |
+
_chem_comp.mon_nstd_parent_comp_id
|
| 39 |
+
_chem_comp.pdbx_synonyms
|
| 40 |
+
_chem_comp.formula_weight
|
| 41 |
+
180-180 180-180 NON-POLYMER ? ? ? ?
|
| 42 |
+
ASP ASP 'L-PEPTIDE LINKING' ? ? ? ?
|
| 43 |
+
ORO ORO NON-POLYMER ? ? ? ?
|
| 44 |
+
ZN ZN NON-POLYMER ? ? ? ?
|
| 45 |
+
#
|
| 46 |
+
loop_
|
| 47 |
+
_chem_comp_bond.comp_id
|
| 48 |
+
_chem_comp_bond.atom_id_1
|
| 49 |
+
_chem_comp_bond.atom_id_2
|
| 50 |
+
_chem_comp_bond.value_order
|
| 51 |
+
_chem_comp_bond.pdbx_aromatic_flag
|
| 52 |
+
ASP N CA SING N
|
| 53 |
+
ASP N H SING N
|
| 54 |
+
ASP N H2 SING N
|
| 55 |
+
ASP CA C SING N
|
| 56 |
+
ASP CA CB SING N
|
| 57 |
+
ASP CA HA SING N
|
| 58 |
+
ASP C O DOUB N
|
| 59 |
+
ASP C OXT SING N
|
| 60 |
+
ASP CB CG SING N
|
| 61 |
+
ASP CB HB2 SING N
|
| 62 |
+
ASP CB HB3 SING N
|
| 63 |
+
ASP CG OD1 DOUB N
|
| 64 |
+
ASP CG OD2 SING N
|
| 65 |
+
ASP OD2 HD2 SING N
|
| 66 |
+
ASP OXT HXT SING N
|
| 67 |
+
ORO N1 C2 SING Y
|
| 68 |
+
ORO N1 C6 SING Y
|
| 69 |
+
ORO N1 HN1 SING N
|
| 70 |
+
ORO C2 O2 DOUB N
|
| 71 |
+
ORO C2 N3 SING Y
|
| 72 |
+
ORO N3 C4 SING Y
|
| 73 |
+
ORO N3 HN3 SING N
|
| 74 |
+
ORO C4 O4 DOUB N
|
| 75 |
+
ORO C4 C5 SING Y
|
| 76 |
+
ORO C5 C6 DOUB Y
|
| 77 |
+
ORO C5 H5 SING N
|
| 78 |
+
ORO C6 C7 SING N
|
| 79 |
+
ORO C7 O71 DOUB N
|
| 80 |
+
ORO C7 O72 SING N
|
| 81 |
+
ORO O72 HO7 SING N
|
| 82 |
+
#
|
| 83 |
+
loop_
|
| 84 |
+
_atom_site.group_PDB
|
| 85 |
+
_atom_site.type_symbol
|
| 86 |
+
_atom_site.label_atom_id
|
| 87 |
+
_atom_site.label_alt_id
|
| 88 |
+
_atom_site.label_comp_id
|
| 89 |
+
_atom_site.label_asym_id
|
| 90 |
+
_atom_site.label_entity_id
|
| 91 |
+
_atom_site.label_seq_id
|
| 92 |
+
_atom_site.pdbx_PDB_ins_code
|
| 93 |
+
_atom_site.auth_seq_id
|
| 94 |
+
_atom_site.auth_comp_id
|
| 95 |
+
_atom_site.auth_asym_id
|
| 96 |
+
_atom_site.auth_atom_id
|
| 97 |
+
_atom_site.B_iso_or_equiv
|
| 98 |
+
_atom_site.occupancy
|
| 99 |
+
_atom_site.pdbx_formal_charge
|
| 100 |
+
_atom_site.annotation_src-component_1_atom
|
| 101 |
+
_atom_site.annotation_expsegmax_1_residue
|
| 102 |
+
_atom_site.mask_expsegmin_1_residue
|
| 103 |
+
_atom_site.condition_hbond_1_atom
|
| 104 |
+
_atom_site.condition_index-group_1_token
|
| 105 |
+
_atom_site.annotation_expsegmin_1_residue
|
| 106 |
+
_atom_site.mask_index_1_residue
|
| 107 |
+
_atom_site.mask_sequence_1_residue
|
| 108 |
+
_atom_site.mask_expsegmax_1_residue
|
| 109 |
+
_atom_site.mask_coordinate_1_atom
|
| 110 |
+
_atom_site.atom_array_index
|
| 111 |
+
_atom_site.Cartn_x
|
| 112 |
+
_atom_site.Cartn_y
|
| 113 |
+
_atom_site.Cartn_z
|
| 114 |
+
_atom_site.pdbx_PDB_model_num
|
| 115 |
+
_atom_site.id
|
| 116 |
+
HETATM X UNK . 180-180 A 0 0 . 0 180-180 A UNK nan 0.0 ? . 180 True 0 0 180 True False True False 0 nan nan nan 1 1
|
| 117 |
+
ATOM C CG . ASP A 0 1 . 1 ASP A CG 17.39 1.0 ? A250 -1 False 0 4 -1 False True False True 1 0.477 -0.542 2.937 1 2
|
| 118 |
+
ATOM O OD1 . ASP A 0 1 . 1 ASP A OD1 18.72 1.0 ? A250 -1 False 0 4 -1 False True False True 2 0.547 0.656 2.665 1 3
|
| 119 |
+
HETATM N N1 . ORO C 2 410 . 410 ORO C N1 18.66 1.0 ? ORO -1 False 0 1 -1 True True False True 3 -2.127 0.345 -1.747 1 4
|
| 120 |
+
HETATM C C2 . ORO C 2 410 . 410 ORO C C2 23.79 1.0 ? ORO -1 False 0 1 -1 True True False True 4 -1.547 -0.999 -1.559 1 5
|
| 121 |
+
HETATM O O2 . ORO C 2 410 . 410 ORO C O2 19.98 1.0 ? ORO -1 False 0 1 -1 True True False False 5 0.0 0.0 0.0 1 6
|
| 122 |
+
HETATM N N3 . ORO C 2 410 . 410 ORO C N3 26.76 1.0 ? ORO -1 False 0 1 -1 True True False True 6 -0.172 -1.019 -1.547 1 7
|
| 123 |
+
HETATM C C4 . ORO C 2 410 . 410 ORO C C4 100.0 1.0 ? ORO -1 False 0 1 -1 True True False True 7 0.602 -0.003 -2.126 1 8
|
| 124 |
+
HETATM O O4 . ORO C 2 410 . 410 ORO C O4 39.79 1.0 ? ORO -1 False 0 1 -1 True True False False 8 0.0 0.0 0.0 1 9
|
| 125 |
+
HETATM C C5 . ORO C 2 410 . 410 ORO C C5 30.44 1.0 ? ORO -1 False 0 1 -1 True True False False 9 0.0 0.0 0.0 1 10
|
| 126 |
+
HETATM C C6 . ORO C 2 410 . 410 ORO C C6 29.02 1.0 ? ORO -1 False 0 1 -1 True True False False 10 0.0 0.0 0.0 1 11
|
| 127 |
+
HETATM C C7 . ORO C 2 410 . 410 ORO C C7 21.02 1.0 ? ORO -1 False 0 1 -1 True True False False 11 0.0 0.0 0.0 1 12
|
| 128 |
+
HETATM O O71 . ORO C 2 410 . 410 ORO C O71 17.46 1.0 ? ORO -1 False 0 1 -1 True True False False 12 0.0 0.0 0.0 1 13
|
| 129 |
+
HETATM O O72 . ORO C 2 410 . 410 ORO C O72 18.24 1.0 ? ORO -1 False 0 1 -1 True True False False 13 0.0 0.0 0.0 1 14
|
| 130 |
+
HETATM ZN ZN . ZN B 1 400 . 400 ZN B ZN 22.02 1.0 +2 ZN -1 False 0 2 -1 True True False True 14 2.223 1.561 1.378 1 15
|
| 131 |
+
#
|
| 132 |
+
_rfd_design.sampled_contig 180-180
|
| 133 |
+
#
|
| 134 |
+
loop_
|
| 135 |
+
_chem_comp_atom.comp_id
|
| 136 |
+
_chem_comp_atom.atom_id
|
| 137 |
+
_chem_comp_atom.type_symbol
|
| 138 |
+
_chem_comp_atom.charge
|
| 139 |
+
_chem_comp_atom.pdbx_aromatic_flag
|
| 140 |
+
_chem_comp_atom.pdbx_leaving_atom_flag
|
| 141 |
+
_chem_comp_atom.pdbx_model_Cartn_x_ideal
|
| 142 |
+
_chem_comp_atom.pdbx_model_Cartn_y_ideal
|
| 143 |
+
_chem_comp_atom.pdbx_model_Cartn_z_ideal
|
| 144 |
+
ASP N N 0 N N -0.31700000166893005 1.687999963760376 0.06599999964237213
|
| 145 |
+
ASP CA C 0 N N -0.4699999988079071 0.28600001335144043 -0.3440000116825104
|
| 146 |
+
ASP C C 0 N N -1.8680000305175781 -0.18000000715255737 -0.028999999165534973
|
| 147 |
+
ASP O O 0 N N -2.5339999198913574 0.41499999165534973 0.7860000133514404
|
| 148 |
+
ASP CB C 0 N N 0.5389999747276306 -0.5799999833106995 0.4129999876022339
|
| 149 |
+
ASP CG C 0 N N 1.937999963760376 -0.19499999284744263 0.004000000189989805
|
| 150 |
+
ASP OD1 O 0 N N 2.1089999675750732 0.6809999942779541 -0.8100000023841858
|
| 151 |
+
ASP OD2 O 0 N N 2.992000102996826 -0.8259999752044678 0.5429999828338623
|
| 152 |
+
ASP OXT O 0 N Y -2.374000072479248 -1.25600004196167 -0.6520000100135803
|
| 153 |
+
ASP H H 0 N N -0.9279999732971191 2.2890000343322754 -0.46700000762939453
|
| 154 |
+
ASP H2 H 0 N Y -0.4779999852180481 1.7949999570846558 1.055999994277954
|
| 155 |
+
ASP HA H 0 N N -0.2919999957084656 0.19900000095367432 -1.4160000085830688
|
| 156 |
+
ASP HB2 H 0 N N 0.4189999997615814 -0.42500001192092896 1.4850000143051147
|
| 157 |
+
ASP HB3 H 0 N N 0.367000013589859 -1.6299999952316284 0.17599999904632568
|
| 158 |
+
ASP HD2 H 0 N N 3.86899995803833 -0.5450000166893005 0.25
|
| 159 |
+
ASP HXT H 0 N Y -3.2750000953674316 -1.5169999599456787 -0.41600000858306885
|
| 160 |
+
ORO N1 N 0 Y N 1.1260000467300415 0.004999999888241291 0.041999999433755875
|
| 161 |
+
ORO C2 C 0 Y N 1.1929999589920044 -0.0010000000474974513 -1.2990000247955322
|
| 162 |
+
ORO O2 O 0 N N 2.2839999198913574 -0.0010000000474974513 -1.8359999656677246
|
| 163 |
+
ORO N3 N 0 Y N 0.08799999952316284 -0.0 -2.062999963760376
|
| 164 |
+
ORO C4 C 0 Y N -1.1349999904632568 -0.0 -1.4950000047683716
|
| 165 |
+
ORO O4 O 0 N N -2.138000011444092 0.0 -2.184000015258789
|
| 166 |
+
ORO C5 C 0 Y N -1.2359999418258667 -0.0 -0.029999999329447746
|
| 167 |
+
ORO C6 C 0 Y N -0.08699999749660492 -0.0 0.6949999928474426
|
| 168 |
+
ORO C7 C 0 N N -0.13899999856948853 -0.0 2.177000045776367
|
| 169 |
+
ORO O71 O 0 N N -1.2109999656677246 -0.0 2.746000051498413
|
| 170 |
+
ORO O72 O 0 N N 1.0019999742507935 -0.0 2.8910000324249268
|
| 171 |
+
ORO HN1 H 0 N N 1.9450000524520874 0.004000000189989805 0.5609999895095825
|
| 172 |
+
ORO HN3 H 0 N N 0.17000000178813934 -0.0 -3.0299999713897705
|
| 173 |
+
ORO H5 H 0 N N -2.1989998817443848 -0.0 0.4569999873638153
|
| 174 |
+
ORO HO7 H 0 N N 0.9679999947547913 -0.0 3.8570001125335693
|
| 175 |
+
ZN ZN ZN 2 N N nan nan nan
|
| 176 |
+
180-180 UNK X 0 N N nan nan nan
|
| 177 |
+
#
|
examples/monomer.cif
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data_monomer-short.cif
|
| 2 |
+
#
|
| 3 |
+
_entry.id monomer-short.cif
|
| 4 |
+
_entry.author rib7
|
| 5 |
+
_entry.date 2026-08-28
|
| 6 |
+
_entry.time 19:44:50
|
| 7 |
+
#
|
| 8 |
+
_entity.id 0
|
| 9 |
+
_entity.type polymer
|
| 10 |
+
_entity.pdbx_description .
|
| 11 |
+
#
|
| 12 |
+
_entity_poly.entity_id 0
|
| 13 |
+
_entity_poly.type polypeptide(L)
|
| 14 |
+
_entity_poly.nstd_linkage no
|
| 15 |
+
_entity_poly.nstd_monomer yes
|
| 16 |
+
_entity_poly.pdbx_seq_one_letter_code X
|
| 17 |
+
_entity_poly.pdbx_seq_one_letter_code_can X
|
| 18 |
+
_entity_poly.pdbx_strand_id A
|
| 19 |
+
_entity_poly.pdbx_target_identifier ?
|
| 20 |
+
#
|
| 21 |
+
_entity_poly_seq.entity_id 0
|
| 22 |
+
_entity_poly_seq.hetero n
|
| 23 |
+
_entity_poly_seq.mon_id 100-100
|
| 24 |
+
_entity_poly_seq.num 1
|
| 25 |
+
#
|
| 26 |
+
_chem_comp.id 100-100
|
| 27 |
+
_chem_comp.name 100-100
|
| 28 |
+
_chem_comp.type NON-POLYMER
|
| 29 |
+
_chem_comp.formula ?
|
| 30 |
+
_chem_comp.mon_nstd_parent_comp_id ?
|
| 31 |
+
_chem_comp.pdbx_synonyms ?
|
| 32 |
+
_chem_comp.formula_weight ?
|
| 33 |
+
#
|
| 34 |
+
_atom_site.group_PDB HETATM
|
| 35 |
+
_atom_site.type_symbol X
|
| 36 |
+
_atom_site.label_atom_id UNK
|
| 37 |
+
_atom_site.label_alt_id .
|
| 38 |
+
_atom_site.label_comp_id 100-100
|
| 39 |
+
_atom_site.label_asym_id A
|
| 40 |
+
_atom_site.label_entity_id 0
|
| 41 |
+
_atom_site.label_seq_id 1
|
| 42 |
+
_atom_site.pdbx_PDB_ins_code .
|
| 43 |
+
_atom_site.auth_seq_id 1
|
| 44 |
+
_atom_site.auth_comp_id 100-100
|
| 45 |
+
_atom_site.auth_asym_id A
|
| 46 |
+
_atom_site.auth_atom_id UNK
|
| 47 |
+
_atom_site.B_iso_or_equiv nan
|
| 48 |
+
_atom_site.occupancy 0.0
|
| 49 |
+
_atom_site.pdbx_formal_charge ?
|
| 50 |
+
_atom_site.annotation_expsegmax_1_residue 100
|
| 51 |
+
_atom_site.mask_expsegmin_1_residue True
|
| 52 |
+
_atom_site.mask_expsegmax_1_residue True
|
| 53 |
+
_atom_site.annotation_expsegmin_1_residue 100
|
| 54 |
+
_atom_site.condition_hbond_1_atom 0
|
| 55 |
+
_atom_site.condition_index-group_1_token 0
|
| 56 |
+
_atom_site.mask_index_1_residue True
|
| 57 |
+
_atom_site.mask_coordinate_1_atom False
|
| 58 |
+
_atom_site.mask_sequence_1_residue False
|
| 59 |
+
_atom_site.annotation_src-component_1_atom .
|
| 60 |
+
_atom_site.atom_array_index 0
|
| 61 |
+
_atom_site.Cartn_x nan
|
| 62 |
+
_atom_site.Cartn_y nan
|
| 63 |
+
_atom_site.Cartn_z nan
|
| 64 |
+
_atom_site.pdbx_PDB_model_num 1
|
| 65 |
+
_atom_site.id 1
|
| 66 |
+
#
|
| 67 |
+
_rfd_design.sampled_contig 100-100
|
| 68 |
+
#
|
| 69 |
+
_chem_comp_atom.comp_id 100-100
|
| 70 |
+
_chem_comp_atom.atom_id UNK
|
| 71 |
+
_chem_comp_atom.type_symbol X
|
| 72 |
+
_chem_comp_atom.charge 0
|
| 73 |
+
_chem_comp_atom.pdbx_aromatic_flag N
|
| 74 |
+
_chem_comp_atom.pdbx_leaving_atom_flag N
|
| 75 |
+
_chem_comp_atom.pdbx_model_Cartn_x_ideal nan
|
| 76 |
+
_chem_comp_atom.pdbx_model_Cartn_y_ideal nan
|
| 77 |
+
_chem_comp_atom.pdbx_model_Cartn_z_ideal nan
|
| 78 |
+
#
|
examples/protein_binder/design.cif
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
examples/protein_binder/input.cif
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
flow_checkpoints/rfd4_proteina_flow-ema.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c1c8b7068e87c41a334ec5b77e7d7cdda30bd31ed4504909d93bd72f7742afe4
|
| 3 |
+
size 1751488703
|
flow_checkpoints/rfd4_proteina_flow-ema.ckpt.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
c1c8b7068e87c41a334ec5b77e7d7cdda30bd31ed4504909d93bd72f7742afe4 rfd4_proteina_flow-ema.ckpt
|
flow_checkpoints/rfd4_proteina_flow_pretrain-ema.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:561bbbe3e97f37ded11ddc934ff2184adcd98038396f7aa1f16e6f9d57f0f8b2
|
| 3 |
+
size 1743470486
|
flow_checkpoints/rfd4_proteina_flow_pretrain-ema.ckpt.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
561bbbe3e97f37ded11ddc934ff2184adcd98038396f7aa1f16e6f9d57f0f8b2 rfd4_proteina_flow_pretrain-ema.ckpt
|
inference.yaml
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Default configs for inference through generate.py
|
| 2 |
+
# The defaults are geared more towards ease-of-use by designers, so running with
|
| 3 |
+
# the existing analysis scripts will necessitate overrides
|
| 4 |
+
|
| 5 |
+
defaults:
|
| 6 |
+
- _self_
|
| 7 |
+
- inference_experiment: inference_on_user_inputs # Default for design
|
| 8 |
+
- recycle_strategy: thinking_medium
|
| 9 |
+
|
| 10 |
+
# NOTE: The experiment configs can override values specified in this config
|
| 11 |
+
|
| 12 |
+
run_name_: inference
|
| 13 |
+
verbose: false # Set true for routine warnings and detailed diagnostic output.
|
| 14 |
+
hf_ckpt_path: nvidia/RFD4-Proteina # Optional local model bundle or HF repo; overrides ckpt_path/name when set.
|
| 15 |
+
hf_ckpt_revision: a06e2509c7033b010fc43614e9fd840303588502 # Pin a Hub commit for reproducible model/AE loading.
|
| 16 |
+
|
| 17 |
+
out_dir: null # If null, defaults to ./runs/<run_name>. For production, set to an external
|
| 18 |
+
# path (e.g. /shared/runs/experiment_name). SLURM scripts set this
|
| 19 |
+
# automatically. All outputs (samples, metrics) go under this directory.
|
| 20 |
+
|
| 21 |
+
# Current SOTA; generation and regression tooling resolve the immutable checkpoint from here.
|
| 22 |
+
# Regenerate the inference baselines when advancing it; use CLI overrides for one-off checkpoints.
|
| 23 |
+
ckpt_name: rfd4-proteina-flow-release-2026-10-19.ckpt
|
| 24 |
+
ckpt_path: ${oc.env:DATA_PATH}/flow_checkpoints
|
| 25 |
+
adapter_path: null # Local PEFT adapter directory; load onto the exact base/AE pair used for fine-tuning.
|
| 26 |
+
|
| 27 |
+
# Only applies when a pre-oracle design filter is active (i.e. its tag gate matched).
|
| 28 |
+
# Determines whether to delete structure files for designs that fail this filter
|
| 29 |
+
delete_failing_pre_oracle: True
|
| 30 |
+
|
| 31 |
+
autoencoder_ckpt_path: ${oc.env:DATA_PATH}/ae_checkpoints/rfd4-proteina-ae-release-2026-10-19.ckpt
|
| 32 |
+
|
| 33 |
+
num_replicates: 1 # Number of times to replicate each example
|
| 34 |
+
|
| 35 |
+
# Token counts to pad batches to for torch.compile. Multiples of 16; null disables.
|
| 36 |
+
compile_lengths: [128, 160, 192, 224, 256, 288, 320, 352, 384, 448, 512, 640, 768, 896, 1024]
|
| 37 |
+
|
| 38 |
+
# Network settings that may change after training without touching the weights; null keeps the checkpoint's
|
| 39 |
+
# value. Example: nn.torch_compile_mode=reduce-overhead nn.tri_mult_backend=cuequivariance
|
| 40 |
+
nn:
|
| 41 |
+
torch_compile_mode: null
|
| 42 |
+
torch_compile_fullgraph: null
|
| 43 |
+
tri_mult_backend: null
|
| 44 |
+
attention_type: null
|
| 45 |
+
|
| 46 |
+
precision: bf16-mixed
|
| 47 |
+
|
| 48 |
+
batch_size: 1
|
| 49 |
+
subset_to_examples: null # If specified, only run inference on these example indices
|
| 50 |
+
prevalidate_inputs: false # Whether to sanity-check each unique input once prior to generation (fail fast on bad inputs).
|
| 51 |
+
seed: null # Will set a random seed if null
|
| 52 |
+
torch_deterministic: false # Locks cuDNN/cuBLAS to deterministic algorithms; significantly reduces throughput
|
| 53 |
+
skip_existing: true # If true, skip examples whose output structure files already exist in the output directory
|
| 54 |
+
save_chain_of_interest: false # If true, write per-design chain-of-interest CIFs (only needed for foldseek clustering).
|
| 55 |
+
compress_outputs: false # Compress persistent structure outputs, including refolds and CIF trajectories.
|
| 56 |
+
metrics_tags: [] # If specified, determines which metrics are computed during inference
|
| 57 |
+
|
| 58 |
+
search:
|
| 59 |
+
checkpoints: null # Set checkpoint policies to enable beam search.
|
| 60 |
+
save_states: false
|
| 61 |
+
resume_from: null
|
| 62 |
+
resume_seed: null
|
| 63 |
+
|
| 64 |
+
num_workers: 4 # Number of dataloader workers
|
| 65 |
+
|
| 66 |
+
hardware:
|
| 67 |
+
accelerator: gpu
|
| 68 |
+
ngpus_per_node_: 1
|
| 69 |
+
nnodes_: 1
|
| 70 |
+
|
| 71 |
+
# Cluster job settings (used by launch_infer_from_local.sh)
|
| 72 |
+
ncpus_: 24 # Number of CPUs per task
|
| 73 |
+
gen_njobs: 1 # Number of parallel generation jobs per config
|
| 74 |
+
eval_njobs: 1 # Number of parallel evaluation jobs per config
|
| 75 |
+
job_id: null # Set by launch_generate.sh; used to disambiguate per-job output files
|
| 76 |
+
|
| 77 |
+
hydra:
|
| 78 |
+
run:
|
| 79 |
+
dir: ./hydra_outputs
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5b549dd61adb2f6c6b5cb95bda13f40f5e31556b4e6c755edc0ca6f011c68fb0
|
| 3 |
+
size 1742660596
|
openfold3/runtime/of3_af3_refolding_env-cuda12.docker.tar
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4723a0cbc1d83810005c54fe3cf3b881e2023d45412e883879f2fab686ab3620
|
| 3 |
+
size 4406498304
|
openfold3/runtime/of3_af3_refolding_env-cuda12.docker.tar.sha256
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dcd4fc6523928a57ee858727438576d44556070b56bac1dec27bd530ffff5ec7
|
| 3 |
+
size 106
|
openfold3/runtime/of3_af3_refolding_env-cuda12.sif
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2b17e5b9262614844c71ce1ddfa1a03ce3534b367b1e919ea41866027087c71
|
| 3 |
+
size 4000178176
|
openfold3/runtime/of3_af3_refolding_env-cuda12.sif.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
c2b17e5b9262614844c71ce1ddfa1a03ce3534b367b1e919ea41866027087c71 of3_af3_refolding_env-cuda12.sif
|
openfold3/runtime/of3_af3_refolding_env-cuda12.sqsh
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6013f17d43d5be70fcad6bb9c674a96803023c418ea0c3dde08f8e1d61ad4719
|
| 3 |
+
size 4000133120
|
openfold3/runtime/of3_af3_refolding_env-cuda12.sqsh.sha256
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
6013f17d43d5be70fcad6bb9c674a96803023c418ea0c3dde08f8e1d61ad4719 of3_af3_refolding_env-cuda12.sqsh
|
openfold3/runtime/of3_af3_refolding_env-cuda13.docker.tar
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0b017dfdf3ea481ad7e56ad8b5fad031f499cb2dc5cafacdeb0a96e84c014fda
|
| 3 |
+
size 3911081472
|
openfold3/runtime/of3_af3_refolding_env-cuda13.docker.tar.sha256
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3990a966c2ac76ec25818a0d8cb8dd30589c6bc550007ebaf34be2735019d58a
|
| 3 |
+
size 106
|
openfold3/runtime/of3_af3_refolding_env-cuda13.sif
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7bbd35478ed5e95131efa7756d03a2e754f3508e603ad08bb8f8db0bd89484a1
|
| 3 |
+
size 3502800896
|