tensorflow / tensorflow/recommenders
Retrieval Task Mixed Precision Dtype Issue
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 2k
- Forks
- 300
- PR merge metrics
- No merged PRs in 30d
Description
For a mixed precision approach, I observed that the query_embeddings for the retrieval task gets cast to the global precision policy while the candidate_embedding preserves its dtype. It causes the linalg computation in the call method in retrieval to error due to different dtypes.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in the retrieval task's call method and reproduce the mixed-precision path described in the issue. Inspect the dtypes of query_embeddings and candidate_embedding before the linalg computation; done means the retrieval computation accepts them without a dtype mismatch error under mixed precision.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, tensorflow
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100