Datasets:
Upload specs/gnn_complexity.yaml with huggingface_hub
Browse files- specs/gnn_complexity.yaml +60 -0
specs/gnn_complexity.yaml
ADDED
|
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Research Spec: GNN Architecture Search for Code Complexity Prediction
|
| 2 |
+
#
|
| 3 |
+
# Automated search across 5 GNN architectures and hyperparameters
|
| 4 |
+
# for predicting Ruby method complexity from AST structure.
|
| 5 |
+
#
|
| 6 |
+
# Launch:
|
| 7 |
+
# ratiocinator research specs/gnn_complexity.yaml
|
| 8 |
+
#
|
| 9 |
+
# The dataset (22K Ruby methods) is in the repo branch.
|
| 10 |
+
# No external data staging needed.
|
| 11 |
+
|
| 12 |
+
# What to research
|
| 13 |
+
topic: "Comparing GNN architectures (GCN, SAGE, GAT, GIN, GraphConv) for predicting Ruby code complexity from Abstract Syntax Trees. The baseline uses GraphSAGE with hidden_dim=64, 3 layers, achieving MAE 4.77 on 22K Ruby methods."
|
| 14 |
+
goal_metric: val_mae
|
| 15 |
+
maximize: false
|
| 16 |
+
|
| 17 |
+
# Target codebase
|
| 18 |
+
repo_url: https://github.com/timlawrenz/jubilant-palm-tree.git
|
| 19 |
+
repo_branch: experiment/ratiocinator-gnn-study
|
| 20 |
+
runner_script: scripts/run_complexity_arm.sh
|
| 21 |
+
|
| 22 |
+
# Infrastructure — models are tiny (~50K params), training is fast
|
| 23 |
+
hardware:
|
| 24 |
+
gpu: "RTX 4090"
|
| 25 |
+
num_gpus: 1
|
| 26 |
+
min_cpu_ram_gb: 32
|
| 27 |
+
min_inet_down: 1000.0
|
| 28 |
+
min_cuda_version: 12.0
|
| 29 |
+
max_dph: 0.40
|
| 30 |
+
disk_gb: 50.0
|
| 31 |
+
image: pytorch/pytorch:2.7.0-cuda12.8-cudnn9-runtime
|
| 32 |
+
|
| 33 |
+
data:
|
| 34 |
+
source: none # Dataset is in the repo branch
|
| 35 |
+
|
| 36 |
+
deps:
|
| 37 |
+
pre_install:
|
| 38 |
+
- "apt-get update -qq && apt-get install -y -qq git-lfs > /dev/null 2>&1 || true"
|
| 39 |
+
- "cd /workspace/experiment && git lfs install && git lfs pull"
|
| 40 |
+
- "pip install torch-geometric torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-2.7.0+cu128.html"
|
| 41 |
+
- "pip install pandas tqdm sentence-transformers nltk scikit-learn numpy"
|
| 42 |
+
requirements: requirements.txt
|
| 43 |
+
exclude_from_requirements:
|
| 44 |
+
- torch
|
| 45 |
+
- torchvision
|
| 46 |
+
- torch_geometric
|
| 47 |
+
verify: "python -c \"import torch_geometric; print(f'PyG {torch_geometric.__version__}')\""
|
| 48 |
+
|
| 49 |
+
metrics:
|
| 50 |
+
protocol: json_line
|
| 51 |
+
json_prefix: "METRICS:"
|
| 52 |
+
|
| 53 |
+
# Budget — small models, fast training (~2-5 min per arm)
|
| 54 |
+
max_iterations: 3
|
| 55 |
+
max_dollars: 15.00
|
| 56 |
+
train_timeout_s: 1200
|
| 57 |
+
download_timeout_s: 600
|
| 58 |
+
|
| 59 |
+
# Output
|
| 60 |
+
paper_title: "What Graph Neural Networks Can and Cannot Learn About Code: A Systematic Empirical Study on Ruby AST Analysis"
|