Port GLiNER2.5 (fastino/gliner2.5-multi-v1) zero-shot NER to vllm.cpp
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 423
- Forks
- 53
- Avg merge
- 20h 26m
- Merged PRs (30d)
- 310
Description
Port the GLiNER2.5 model (DeBERTa v2 encoder + disentangled attention + GLiNER2 pooler head) into vllm.cpp as the first encoder-only (BERT-class) model. vLLM has no DeBERTa support; the vllm-factory plugin (ddickmann/vllm-factory) is the reference oracle.
Scope: DeBERTa v2 encoder, GLiNER2 pooler head (SpanRep, CountLSTM, classifier), model registration, OpenAI-compatible serving, CPU + GPU (CUDA + ROCm) support, LocalAI backend.
Local issue: ISSUE-LOCAL-01M2TMQF230HKCX03RW6ADCVCT
Row: MODEL-GLINER25
Spec: .agents/specs/gliner2.5.md
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 with .agents/specs/gliner2.5.md and compare the required DeBERTa v2 encoder and GLiNER2 pooler components against the vllm-factory reference oracle. Then trace model registration, OpenAI-compatible serving, and the LocalAI backend; done means the listed CPU, CUDA, ROCm, serving, and registration scope is supported.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- api, backend-api-design, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100