microsoft / microsoft/TransformerCompression
NaN weights after rotating and slicing
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 460
- Forks
- 58
- Avg merge
- 8h 40m
- Merged PRs (30d)
- 1
Description
Hi!
The state dictionary has weights as shown post slicing and rotating. Are we initializing layers pre/post slicing?
Contributor guide
No contributing guide indexed for this repository
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
The issue names no files or tests; start by reproducing the rotating-and-slicing sequence and inspecting the state dictionary before and after those operations. Trace layer initialization relative to slicing and rotation, and consider the issue done when the cause of the NaN weights is identified and a regression check demonstrates finite weights.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100