# Research Spec: GNN Decoder Topology — Track 4
#
# Tests tree-aware decoding: giving the decoder the actual AST edge structure
# instead of sequential chain edges.  Three modes:
#   - chain:          Legacy baseline (0→1→2→…)
#   - teacher_forced: Ground-truth AST edges during GNN message passing
#   - iterative:      Two-pass: chain→predict parents→rebuild tree→refine
#
# The hypothesis: the decoder's GNN never sees tree topology, so it cannot
# learn structure-sensitive generation.  Providing real edges should improve
# parent prediction accuracy and node-type diversity.
#
# Launch:
#   ratiocinator research specs/gnn_topology.yaml

# What to research
topic: "Decoder topology for GNN code generation: does giving the decoder the real AST tree structure (instead of sequential chain edges) improve reconstruction?  Compare chain baseline, teacher-forced ground-truth edges, and iterative predict-then-refine.  Cross with GAT/GCN/GIN decoder architectures and improved/comprehensive loss functions."
goal_metric: syntactic_validity_pct
maximize: true

# Target codebase
repo_url: https://github.com/timlawrenz/jubilant-palm-tree.git
repo_branch: experiment/ratiocinator-gnn-study
runner_script: scripts/run_topology_arm.sh

# Infrastructure — ~850K params, moderate training
hardware:
  gpu: "RTX 4090"
  num_gpus: 1
  min_cpu_ram_gb: 32
  min_inet_down: 1000.0
  min_cuda_version: 12.0
  max_dph: 0.40
  disk_gb: 50.0
  image: pytorch/pytorch:2.7.0-cuda12.8-cudnn9-runtime

data:
  source: none  # Dataset is in the repo branch

deps:
  pre_install:
    - "apt-get update -qq && apt-get install -y -qq git-lfs > /dev/null 2>&1 || true"
    - "cd /workspace/experiment && git lfs install && git lfs pull"
    - "pip install torch-geometric torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-2.7.0+cu128.html"
    - "pip install pandas tqdm sentence-transformers nltk scikit-learn numpy"
  requirements: requirements.txt
  exclude_from_requirements:
    - torch
    - torchvision
    - torch_geometric
  verify: "python -c \"import torch_geometric; print(f'PyG {torch_geometric.__version__}')\""

metrics:
  protocol: json_line
  json_prefix: "METRICS:"

# Budget — ~9 arms (3 edge modes × 3 conv types), ~10 min each
max_iterations: 2
max_dollars: 15.00
train_timeout_s: 2400
download_timeout_s: 600

# Output
paper_title: "What Graph Neural Networks Can and Cannot Learn About Code: A Systematic Empirical Study on Ruby AST Analysis"