Datasetpublic

timlawrenz/gnn-ruby-code-study

GNN Ruby Code Study Systematic study of Graph Neural Network architectures for Ruby code complexity prediction and generation. Paper: Graph Neural Networks for Ruby Code Complexity Prediction and Generation: A Systematic Architecture Study Dataset 22,452 Ruby methods parsed into AST graphs with 74-dimensional node features. Split Samples File Train 19,084 dataset/train.jsonl Validation 3,368 dataset/val.jsonl Each JSONL record contains:… See the full description on the dataset page: https://huggingface.co/datasets/timlawrenz/gnn-ruby-code-study.

sourceHugging Facemitupdated 6mo agoView on Hugging Face

0likes101downloads

rootexperimentsgnn_architecture_comparison.yaml

gnn_architecture_comparison.yaml144 linesDownload Raw Back to experiments

1# Fleet Spec: Controlled GNN Architecture Comparison2#3# Direct comparison of 5 GNN architectures (GCN, SAGE, GAT, GIN, GraphConv)4# with controlled hyperparameters for Ruby code complexity prediction.5#6# Launch:7#   ratiocinator fleet run experiments/gnn_architecture_comparison.yaml8#9# Each arm varies ONLY the conv_type. All other hyperparameters are identical.10 11name: gnn-architecture-comparison12description: "5-way GNN architecture comparison on Ruby AST complexity prediction"13 14hardware:15  gpu: "RTX 4090"16  num_gpus: 117  min_cpu_ram_gb: 3218  min_inet_down: 1000.019  min_cuda_version: 12.020  max_dph: 0.4021  disk_gb: 50.022  image: pytorch/pytorch:2.7.0-cuda12.8-cudnn9-runtime23 24repo:25  url: https://github.com/timlawrenz/jubilant-palm-tree.git26  branch: experiment/ratiocinator-gnn-study27  clone_depth: 128 29data:30  source: none31 32deps:33  pre_install:34    - "apt-get update -qq && apt-get install -y -qq git-lfs > /dev/null 2>&1 || true"35    - "cd /workspace/experiment && git lfs install && git lfs pull"36    - "pip install torch-geometric torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-2.7.0+cu128.html"37    - "pip install pandas tqdm sentence-transformers nltk scikit-learn numpy"38  requirements: requirements.txt39  exclude_from_requirements:40    - torch41    - torchvision42    - torch_geometric43  verify: "python -c \"import torch_geometric; print(f'PyG {torch_geometric.__version__}')\""44 45arms:46  # ── Architecture comparison (same hyperparams, different conv) ──47  - name: sage-baseline48    description: "GraphSAGE baseline (original architecture)"49    command: "bash scripts/run_complexity_arm.sh"50    env:51      CONV_TYPE: "SAGE"52      HIDDEN_DIM: "64"53      NUM_LAYERS: "3"54      DROPOUT: "0.1"55      LEARNING_RATE: "0.001"56      EPOCHS: "50"57 58  - name: gcn59    description: "Graph Convolutional Network"60    command: "bash scripts/run_complexity_arm.sh"61    env:62      CONV_TYPE: "GCN"63      HIDDEN_DIM: "64"64      NUM_LAYERS: "3"65      DROPOUT: "0.1"66      LEARNING_RATE: "0.001"67      EPOCHS: "50"68 69  - name: gat70    description: "Graph Attention Network"71    command: "bash scripts/run_complexity_arm.sh"72    env:73      CONV_TYPE: "GAT"74      HIDDEN_DIM: "64"75      NUM_LAYERS: "3"76      DROPOUT: "0.1"77      LEARNING_RATE: "0.001"78      EPOCHS: "50"79 80  - name: gin81    description: "Graph Isomorphism Network"82    command: "bash scripts/run_complexity_arm.sh"83    env:84      CONV_TYPE: "GIN"85      HIDDEN_DIM: "64"86      NUM_LAYERS: "3"87      DROPOUT: "0.1"88      LEARNING_RATE: "0.001"89      EPOCHS: "50"90 91  - name: graphconv92    description: "GraphConv (Morris et al.)"93    command: "bash scripts/run_complexity_arm.sh"94    env:95      CONV_TYPE: "GraphConv"96      HIDDEN_DIM: "64"97      NUM_LAYERS: "3"98      DROPOUT: "0.1"99      LEARNING_RATE: "0.001"100      EPOCHS: "50"101 102  # ── Hyperparameter variants on best-expected architectures ──103  - name: sage-wide104    description: "SAGE with 128 hidden dim"105    command: "bash scripts/run_complexity_arm.sh"106    env:107      CONV_TYPE: "SAGE"108      HIDDEN_DIM: "128"109      NUM_LAYERS: "3"110      DROPOUT: "0.1"111      LEARNING_RATE: "0.001"112      EPOCHS: "50"113 114  - name: gat-wide115    description: "GAT with 128 hidden dim"116    command: "bash scripts/run_complexity_arm.sh"117    env:118      CONV_TYPE: "GAT"119      HIDDEN_DIM: "128"120      NUM_LAYERS: "3"121      DROPOUT: "0.1"122      LEARNING_RATE: "0.001"123      EPOCHS: "50"124 125  - name: sage-deep126    description: "SAGE with 5 layers"127    command: "bash scripts/run_complexity_arm.sh"128    env:129      CONV_TYPE: "SAGE"130      HIDDEN_DIM: "64"131      NUM_LAYERS: "5"132      DROPOUT: "0.1"133      LEARNING_RATE: "0.001"134      EPOCHS: "50"135 136metrics:137  protocol: json_line138  json_prefix: "METRICS:"139 140budget:141  max_dollars: 10.00142  train_timeout_s: 1200143  download_timeout_s: 600144