Fix GraphEdgeActionMLP backward policy bugs#499
Draft
josephdviviano wants to merge 1 commit intomasterfrom
Draft
Fix GraphEdgeActionMLP backward policy bugs#499josephdviviano wants to merge 1 commit intomasterfrom
josephdviviano wants to merge 1 commit intomasterfrom
Conversation
Two bugs in GraphEdgeActionMLP.forward() when is_backward=True: 1. node_class_logits was only assigned in the forward branch but referenced unconditionally in the return TensorDict, causing UnboundLocalError. Fix: compute node_class_logits and edge_class_logits unconditionally before the if/else branch, matching GraphActionGNN pattern. 2. states_tensor.x (shape: total_nodes, 1) was passed directly to nn.Embedding without squeezing, producing 3D output that propagated through pad_sequence into 3D node_index_logits. Fix: squeeze x before embedding, matching GraphActionGNN. Tests added: - test_graph_module_output_shapes (8 parametrized cases) - test_graph_module_output_shapes_on_empty_graphs (4 cases) - test_graph_tb_pipeline (4 cases) - test_graph_ring_smoke parametrized with use_gnn=[True, False] Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## master #499 +/- ##
==========================================
+ Coverage 72.48% 73.20% +0.72%
==========================================
Files 55 55
Lines 8519 8518 -1
Branches 1090 1090
==========================================
+ Hits 6175 6236 +61
+ Misses 1957 1896 -61
+ Partials 387 386 -1
🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
GraphEdgeActionMLP.forward()whenis_backward=Truethat caused crashes during TB loss computation withrecalculate_all_logprobs=Truenode_class_logitswas only assigned in the forward branch but referenced unconditionally. Fixed by computing class logits unconditionally (matchingGraphActionGNNpattern).states_tensor.x(shape(N, 1)) was not squeezed beforenn.Embedding, producing 3Dnode_index_logits. Fixed by adding.squeeze(-1)(matchingGraphActionGNN).Tests added
test_graph_module_output_shapes: 8 parametrized cases (GNN/MLP x forward/backward x directed/undirected)test_graph_module_output_shapes_on_empty_graphs: 4 cases for initial (empty) statestest_graph_tb_pipeline: 4 cases running a full TB training step withrecalculate_all_logprobs=Truetest_graph_ring_smoke: parametrized withuse_gnn=[True, False]Test plan
🤖 Generated with Claude Code