Spaces:
Running
Running
File size: 2,826 Bytes
190bb48 e83a069 5d33755 e83a069 5d33755 e83a069 190bb48 e83a069 190bb48 5d33755 190bb48 | 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 | const fs = require('fs');
const vm = require('vm');
const context = {
TDGBENCH_FEATURES:{showKggen:true},
DATASETS_METADATA:{domains:{
biomed:{graphs:{cg:{},GT2KG_tdg:{},kggen:{}}},
dbpedia:{graphs:{cg:{},GT2KG_tdg:{}}}
}}
};
vm.createContext(context);
vm.runInContext(fs.readFileSync('assets/js/pages/evaluate-gnn.js','utf8'),context);
function state(overrides){
return Object.assign({
dataset:'biomed',graph:'GT2KG_tdg',variant:'default',modelSource:'existing',
existingModel:'GCN',pygModel:'SAGEConv',customModule:'my_model',customSymbol:'MyGNN',
emb:'sentence-transformers/all-MiniLM-L6-v2',results:'tdgbench_results'
},overrides||{});
}
let code=context.genC2(state({modelSource:'pyg',pygModel:'SAGEConv'}));
if(!code.includes('from torch_geometric.nn import SAGEConv'))throw new Error('Import PyG absent.');
if(!code.includes('model=SAGEConv'))throw new Error('La convolution PyG doit être passée comme classe.');
code=context.genC2(state({modelSource:'existing',existingModel:'GCN'}));
if(!code.includes('model="GCN"')||code.includes('torch_geometric'))throw new Error('Modèle TDGBench incorrect.');
code=context.genC2(state({modelSource:'custom',customModule:'models',customSymbol:'ResearchGNN'}));
if(!code.includes('from models import ResearchGNN')||!code.includes('model=ResearchGNN'))throw new Error('Modèle personnalisé incorrect.');
code=context.genC2(state({modelSource:'existing',existingModel:'MLP'}));
if(code.includes(' graph=')||!code.includes('model="MLP"'))throw new Error('Baseline MLP incorrecte.');
code=context.genC2(state({dataset:'dbpedia',graph:'all',modelSource:'pyg',pygModel:'SAGEConv'}));
if(code.includes('available_graphs')||code.includes('for graph in graphs'))throw new Error('Tous les graphes doivent produire des appels explicites.');
if(!code.includes('tdg = bench.unified_evaluation(')||!code.includes('graph="GT2KG_tdg"'))throw new Error('L\'évaluation explicite du TDG DBpedia est absente.');
if(!code.includes('cg = bench.unified_evaluation(')||!code.includes('graph="cg"'))throw new Error('L\'évaluation explicite du graphe curé DBpedia est absente.');
if(!code.includes('print(tdg.best_results().to_string(index=False))')||!code.includes('print(cg.best_results().to_string(index=False))'))throw new Error('Les meilleurs résultats des deux graphes ne sont pas affichés.');
if((code.match(/model=SAGEConv/g)||[]).length!==2)throw new Error('SAGEConv doit être transmis aux deux évaluations.');
code=context.genC2(state({dataset:'biomed',graph:'all',modelSource:'pyg',pygModel:'SAGEConv'}));
if(!code.includes('graph="kggen"'))throw new Error('KGGen doit être inclus dans l’évaluation de tous les graphes biomédicaux.');
console.log('Evaluation GNN valide : TDGBench, PyG, modèle personnalisé, MLP et tous les graphes.');
|