numind/NuNER_Zero
Primitive: /extract · Extract ·
DeBERTa
NuNER Zero is a zero-shot Named Entity Recognition (NER) Model. (Check NuNER for the few-shot setting).
Entities
Overview
Hardware: — drives latency, throughput & cost
| Size | 449M params |
|---|---|
| Tasks | /extract |
| License | mit |
| Languages | en |
| Latency | 107 ms |
| Throughput | 14.9K tok/s |
| Cost | $0.015 /1M tok |
Cost is approximate — computed from list GPU prices; your actual price depends on the provider you deploy SIE with.
Extraction
| Output kinds | Entities |
|---|---|
| Inputs | text |
| Max sequence length | 384 |
Benchmarks
CoNLL-2003
Named entity recognition on Reuters newswire text
Corpus: 3,453 Queries: 3,453
Quality
f1 0.6068
precision 0.5430
recall 0.6875
Performance L4 b1 c16
Extract 7.1K tok/s
Extract p50 97.0ms
SPIA
Personally identifiable information (PII) span detection for redaction workflows
Corpus: 151 Queries: 151
Quality
f1 0.4632
precision 0.4727
recall 0.4540
Performance L4 b1 c16
Extract 22.6K tok/s
Extract p50 116.1ms
Compare (0)Compare →