Vela Domain
Vela Domain identifies the subject of a request across 14 domains to help select relevant models and tools.
307M parameters · Input capacity: 32,768 tokens, including special tokens.
Evaluation
Macro F1 (×100) on the same development sets, compared with the original mmBERT32K Domain model. Higher is better.
| Evaluation | Original mmBERT | Vela |
|---|---|---|
| Short requests · 1,988 | 66.33 | 85.15 |
| Controlled 4K context · 28 | 7.14 | 71.19 |
| Controlled 8K context · 28 | 1.79 | 81.19 |
| Controlled 16K context · 28 | 1.79 | 81.19 |
| Controlled 32K context · 28 | 1.79 | 77.86 |
Short requests cover English, Chinese, Spanish, French, German and Japanese. Long-context evaluation uses 28 English/Chinese examples extended to each length. Both models process the complete input in FP32. These development sets informed Vela checkpoint selection; they are not an independent test set.
Quick start
With PyTorch and Transformers 4.57.6 or 5.17.0:
from transformers import pipeline
model_id = "llm-semantic-router/Vela-1.0-Encoder-307M-Domain"
model = pipeline("text-classification", model=model_id, device=-1)
print(model("Why do bond prices fall when interest rates rise?", top_k=None, truncation=False))