Activities in Train GGNN
4 activities · 1 phase
| # | Name | Abbrev. | Guidance | Dependencies | Actions |
|---|---|---|---|---|---|
|
Model Training
4
|
|||||
| 1 | Batch graphs as disconnected components | TGN-1 | Represent each graph as a sparse adjacency list to minimize memory. Combine a minibatch into … | ||
| 1 | Configure task-specific graph | TGN-1 | For VARNAMING: replace all tokens of the target variable with a special SLOT token. For … | ||
| 1 | Run GGNN propagation | TGN-1 | For each step: compute messages m(v)_k = linear(h(v)) per edge type. Aggregate incoming messages via … | ||
| 1 | Train with maximum likelihood | TGN-1 | VARNAMING: average final SLOT node states; use as GRU initial state to predict variable name … | ||