Text Classification
Transformers
ONNX
Safetensors
modernbert
semantic-router
vela
text-embeddings-inference
Instructions to use vllm-sr/Vela-1.0-Encoder-307M-FactCheck with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use vllm-sr/Vela-1.0-Encoder-307M-FactCheck with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="vllm-sr/Vela-1.0-Encoder-307M-FactCheck")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForSequenceClassification tokenizer = AutoTokenizer.from_pretrained("vllm-sr/Vela-1.0-Encoder-307M-FactCheck") model = AutoModelForSequenceClassification.from_pretrained("vllm-sr/Vela-1.0-Encoder-307M-FactCheck", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Commit History
27 days ago
-
Update Hugging Face org references: llm-semantic-router → vllm-sr
Verified signaturec2219a3
-
Add original mmBERT evaluation comparison for Vela FactCheck
Verified signature9d3130e -
Simplify Vela quick starts and model cards
Verified signatured7e9d8e -
Release Vela on the unified Encoder base
Verified signature36c413f -
Unify Vela model cards with Semantic Router branding
Verified signaturee84e690
-
Add Vela Embedding to the model family
Verified signaturee486953 -
Add Vela Feedback to the model family
Verified signaturedce96a8 -
Add Vela Reranker to the model family
Verified signature61901ed -
Introduce the Vela model family and simplify the model card
Verified signaturedf78272 -
Correct Router threshold defaults and preserve fixed evaluation evidence
Verified signature83b0295 -
Release Vela 1.0 Encoder 307M FactCheck with frozen quality and engine evidence
Verified signatureNew text classifier predicting the need for external verification. Includes native FP32 and validated ONNX variants, source-pinned data and training reproduction, independent FP32/BF16 evaluation, and development-selected threshold 0.85. Model card reports increased verification recall, additional false positives and downstream cost, German regression, and limited 32K stress accuracy.018af81