Dependency rankings

Compare AI / ML libraries and shared infrastructure.

Browse AI & ML

Ranked by distinct dependent records in the static graph, not execution frequency. Environments and installed versions stay separate.

Download rankings CSV

ML / AI dependencies

Frameworks, classical ML, native ML libraries, and supporting ML/GPU packages with recorded dependents.

ML and AI dependencies. Counts of distinct dependent records, not runtime usage frequency.
DependencyDirectIncl. indirect
scikit-learn1.4.2 · CoreML library
Inspect 11 direct dependents

2 metadata-declared; 11 linked by static import, native link, or asset evidence. These sets may overlap.

1112
scikit-learn1.6.1 · pandda2ML library
Inspect 4 direct dependents

4 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

44
PyTorch (torch)1.13.1 · pandda2AI / ML framework
Inspect 3 direct dependents

3 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

33
nvidia-cublas-cu1111.10.3.66 · pandda2GPU support
Inspect 2 direct dependents

2 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

25
nvidia-cuda-nvrtc-cu1111.7.99 · pandda2GPU support
Inspect 1 direct dependents

1 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

14
nvidia-cuda-runtime-cu1111.7.99 · pandda2GPU support
Inspect 1 direct dependents

1 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

14
nvidia-cudnn-cu118.5.0.96 · pandda2GPU support
Inspect 1 direct dependents

1 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

14
fms_nnmodel.soVersion not recorded · CoreNative ML
Inspect 1 direct dependents

0 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

13
doublehelix_nnmodel.soVersion not recorded · CoreNative ML
Inspect 1 direct dependents

0 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

12
ONNX Runtime1.19.2 · CoreAI / ML framework
Inspect 1 direct dependents

0 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

12
optuna4.5.0 · CoreML library
Inspect 1 direct dependents

0 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

12
torchmetrics1.5.2 · pandda2ML library
Inspect 1 direct dependents

1 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

12
cbuccaneerNNVersion not recorded · CoreNative ML
Inspect 1 direct dependents

0 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

11
phasertng_ext.soVersion not recorded · CoreNative ML
Inspect 1 direct dependents

0 metadata-declared; 1 linked by static import, native link, or asset evidence. These sets may overlap.

11
hdbscan0.8.42 · pandda2ML library
Inspect 1 direct dependents

1 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

11

Showing 15 of 18 dependencies

Most shared dependencies

Shipped software with the most incoming dependencies. Unresolved JavaScript declarations are excluded; native linkage coverage is partial.

Most shared dependencies. Counts of distinct dependent records, not runtime usage frequency.
DependencyDirectIncl. indirect
numpy1.26.4 · Core
Inspect 33 direct dependents

31 metadata-declared; 2 linked by static import, native link, or asset evidence. These sets may overlap.

3345
numpy1.26.4 · pandda2
Inspect 17 direct dependents

17 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

1718
packaging24.2 · Core
Inspect 15 direct dependents

15 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

1532
libgcc16.2.0 · abcfold
Inspect 14 direct dependents

14 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

1418
libgcc16.2.0 · pandda2
Inspect 13 direct dependents

13 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

1325
python3.9.23 · pandda2
Inspect 12 direct dependents

12 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

1212
scipy1.13.1 · Core
Inspect 11 direct dependents

11 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

1121
scikit-learn1.4.2 · CoreML library
Inspect 11 direct dependents

2 metadata-declared; 11 linked by static import, native link, or asset evidence. These sets may overlap.

1112
six1.12.0 · Core
Inspect 9 direct dependents

9 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

924
typing-extensions4.14.1 · pandda2
Inspect 9 direct dependents

9 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

917
packaging26.3 · pandda2
Inspect 9 direct dependents

9 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

911
pandas2.1.4 · Core
Inspect 9 direct dependents

9 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

910
gemmi0.7.5 · Core
Inspect 8 direct dependents

8 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

88
setuptools79.0.1 · Core
Inspect 7 direct dependents

7 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

717
requests2.28.2 · Core
Inspect 7 direct dependents

7 metadata-declared; 0 linked by static import, native link, or asset evidence. These sets may overlap.

78

Showing 15 of 259 dependencies

Largest recorded ML-linked assets

File sizes, not total framework footprints. These assets have a static runtime-scope relationship to an ML-connected record; some are examples or optional third-party payloads.

Asset / consumerMiBSame-hash paths
dense_weight_0.binphasertng matthews ml · Core
File path and exact size

lib/python3.11/site-packages/phasertng/tensorflow/mle_export/dense_weight_0.bin
333,971,456 bytes

318.51
nucleofind-core.onnxnucleofind · Core
File path and exact size

lib/data/nucleofind_models/nucleofind-core.onnx
148,630,770 bytes

141.72
nucleofind-core.onnxnucleofind · Core
File path and exact size

lib/python3.11/site-packages/nucleofind_models/nucleofind-core.onnx
148,630,770 bytes

141.72
model_event.ckptpandda-2-gemmi · pandda2
File path and exact size

share/mamba/envs/pandda2/lib/python3.9/site-packages/pandda_gemmi/cnn/model_event.ckpt
147,133,214 bytes

140.31
model_build.ckptpandda-2-gemmi · pandda2
File path and exact size

share/mamba/envs/pandda2/lib/python3.9/site-packages/pandda_gemmi/cnn/model_build.ckpt
57,114,976 bytes

54.51
buccaneer_nn_weights.csvbuccaneer nn · Core
File path and exact size

lib/data/buccaneer_nn_weights.csv
32,954,882 bytes

31.41
conv3d_2_weight_0.binphasertng matthews ml · Core
File path and exact size

lib/python3.11/site-packages/phasertng/tensorflow/mle_export/conv3d_2_weight_0.bin
884,736 bytes

0.81
conv3d_1_weight_0.binphasertng matthews ml · Core
File path and exact size

lib/python3.11/site-packages/phasertng/tensorflow/mle_export/conv3d_1_weight_0.bin
221,184 bytes

0.21
trained_classifier.joblibconkit · Core
File path and exact size

lib/python3.11/site-packages/conkit/misc/trained_classifier.joblib
152,258 bytes

0.11
dense_1_weight_0.binphasertng matthews ml · Core
File path and exact size

lib/python3.11/site-packages/phasertng/tensorflow/mle_export/dense_1_weight_0.bin
12,288 bytes

01

Showing 10 of 26 asset paths

Each row is one recorded path. NucleoFind’s 141.7 MiB ONNX file appears twice with the same SHA-256: two shipped paths, one unique payload. MiB = 1,048,576 bytes.

How to read the counts

Direct counts distinct incoming source records. Including indirect counts all distinct upstream records reachable through the same graph; it already includes direct dependents. Cycles and duplicate edges do not inflate either count.

Only runtime-scope requires, conda_depends, imports, links, loads_asset and bundles_asset edges qualify, with both endpoints shipped in the same environment. This retains metadata declarations and cited static links; it does not prove a dependency ran.

Build, optional, external, provenance and ownership-only edges are excluded. In particular, unresolved JavaScript dependency declarations do not count as installed shared dependencies. This is a ranking of the recorded graph, not a complete native linkage census or a count of unique applications. Python and Conda records may overlap.

ML membership uses the CSV’s ml_family field on shipped software, not keyword matching. Package byte sizes are absent from the CSV; only the recorded asset sizes above can be compared.

Source: nodes.csv and edges.csv; ml_reach.csv identifies the connected consumers for the asset table. 580 qualifying edge rows, 570 unique source–target pairs, 259 software dependencies with incoming links.