linkedin / linkedin/Liger-Kernel
[feat] Add support for encoder-only transformers (e.g. BERT)
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 6.6k
- Forks
- 603
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 47
Description
### 🚀 The feature, motivation and pitch
Liger Kernel is currently incompatible with encoder-only transformer architectures such as BERT, DistilBERT, RoBERTa, XLM-R, and DeBERTa.
Given the importance these models still have in research and industry use-cases, it would be great to see support added to further decrease memory requirements and increase training throughput.
### Alternatives
_No response_
### Additional context
_No response_
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
The issue names no files, tests, or entry points. Start by tracing how Liger Kernel currently supports decoder-style transformer architectures, then compare the requirements of BERT, DistilBERT, RoBERTa, XLM-R, and DeBERTa. Done means encoder-only models are supported with reduced memory requirements and improved training throughput.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning, performance
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 30/100