Datasets:
File size: 2,217 Bytes
731072c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 | # Research Spec: GNN Code Generation Failure Analysis
#
# Systematic investigation of why GNNs fail at code generation.
# Tests loss functions, decoder architectures, embedding dimensions.
#
# Launch:
# ratiocinator research specs/gnn_generation.yaml
#
# Requires pre-collated data (.pt files). Either stage via rsync
# or generate on-instance from the JSONL files in the branch.
# What to research
topic: "Systematic analysis of why Graph Neural Networks fail at code generation from ASTs. Testing hypotheses: (1) MSE vs cross-entropy loss on node types, (2) different decoder GNN architectures (GCN/GAT/GIN/SAGE), (3) embedding dimension (64/128/256). Baseline: 0% syntactic validity with GAT decoder, MSE loss, 64D embeddings."
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_generation_arm.sh
# Infrastructure — decoder models ~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; pre-collation done on-instance
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 — longer training per arm (~10-20 min)
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"
|