Training with ConvE and GNN Encoder Disabled fails with standard params

Forschungsprojekt-Informatik / CompGCN

ICLR 2020: Composition-Based Multi-Relational Graph Convolutional Networks

Apache License 2.0

0 stars 0 forks source link

Training with ConvE and GNN Encoder Disabled fails with standard params #14

Open A7exSchin opened 1 year ago

A7exSchin commented 1 year ago

Executing python run.py -score_func conve -opn sub -ker_sz 5 -data codex-s -name ConvE_Sub_codex_s_no_GCN -disable_gnn_encoder true

fails with error RuntimeError: mat1 and mat2 shapes cannot be multiplied (64x200 and 100x2034).

This can be fixed by setting the parameter -k_h 10. This should be documented, which numbers are to choose in which situation.

1Maxnet1 commented 1 year ago

Did we make any process here? Do we have documented somewhere which parameters work for which model or do we only get them from the logs in the result repository?