Position embeddings allocated but never used - model is bag-of-tokens #1

Open

opened

Found via code audit. transformer.py:52. x = tok_emb directly without adding position_embedding. Model cannot learn sequence order. All training produces garbage.

to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Assignees

No one assigned

Labels

No labels

No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests