experiments/gnn_generation_analysis.yaml

# Fleet Spec: GNN Code Generation Failure Analysis (Track 2)
#
# Systematic analysis of code generation quality across decoder configurations.
# Tests: loss functions × decoder architectures × hidden dimensions.
#
# Launch:
#   ratiocinator fleet run experiments/gnn_generation_analysis.yaml

name: gnn-generation-analysis
description: "Systematic GNN code generation failure analysis"

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

repo:
  url: https://github.com/timlawrenz/jubilant-palm-tree.git
  branch: experiment/ratiocinator-gnn-study
  clone_depth: 1

data:
  source: none

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__}')\""

arms:
  # ── Loss function comparison (GAT decoder baseline) ──
  - name: improved-loss-gat
    description: "Improved (cross-entropy) loss, GAT decoder"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "GAT"
      LOSS_FN: "improved"
      HIDDEN_DIM: "256"
      NUM_LAYERS: "5"
      EPOCHS: "30"

  - name: simple-loss-gat
    description: "Simple (MSE) loss, GAT decoder"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "GAT"
      LOSS_FN: "simple"
      HIDDEN_DIM: "256"
      NUM_LAYERS: "5"
      EPOCHS: "30"

  - name: comprehensive-loss-gat
    description: "Comprehensive (combined) loss, GAT decoder"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "GAT"
      LOSS_FN: "comprehensive"
      HIDDEN_DIM: "256"
      NUM_LAYERS: "5"
      EPOCHS: "30"

  # ── Decoder architecture comparison (improved loss) ──
  - name: improved-loss-sage
    description: "Improved loss, SAGE decoder"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "SAGE"
      LOSS_FN: "improved"
      HIDDEN_DIM: "256"
      NUM_LAYERS: "5"
      EPOCHS: "30"

  - name: improved-loss-gin
    description: "Improved loss, GIN decoder"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "GIN"
      LOSS_FN: "improved"
      HIDDEN_DIM: "256"
      NUM_LAYERS: "5"
      EPOCHS: "30"

  - name: improved-loss-gcn
    description: "Improved loss, GCN decoder"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "GCN"
      LOSS_FN: "improved"
      HIDDEN_DIM: "256"
      NUM_LAYERS: "5"
      EPOCHS: "30"

  # ── Wider decoder ──
  - name: improved-loss-gat-wide
    description: "Improved loss, GAT decoder, hidden_dim=512"
    command: "bash scripts/run_generation_arm.sh"
    env:
      DECODER_CONV_TYPE: "GAT"
      LOSS_FN: "improved"
      HIDDEN_DIM: "512"
      NUM_LAYERS: "5"
      EPOCHS: "30"

metrics:
  protocol: json_line
  json_prefix: "METRICS:"

budget:
  max_dollars: 10.00
  train_timeout_s: 3600
  download_timeout_s: 600