Instructions to use LaraAI-Labs/KURE-v1-onnx-int8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- sentence-transformers
How to use LaraAI-Labs/KURE-v1-onnx-int8 with sentence-transformers:
from sentence_transformers import SentenceTransformer model = SentenceTransformer("LaraAI-Labs/KURE-v1-onnx-int8") sentences = [ "The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium." ] embeddings = model.encode(sentences) similarities = model.similarity(embeddings, embeddings) print(similarities.shape) # [3, 3] - Notebooks
- Google Colab
- Kaggle
KURE-v1 ONNX (INT8)
nlpai-lab/KURE-v1 (BGE-M3 νκ΅μ΄ νμΈνλ)μ
ONNX INT8 λ³νλ³Έ. μ¨λλ°μ΄μ€(μ£μ§) νκ΅μ΄ μλ―Έκ²μ μλ² λμ©.
μ€ν
- μμν: int8 λμ μμν κ°μ€μΉ + fp32 μΆλ ₯ (
onnxruntime.quantization.quantize_dynamic, QInt8). ~568MB. - κ·Έλν: opset 17, XLM-RoBERTa.
- μ
λ ₯:
input_ids,attention_mask(2-input, λμ batch/seq). μΆλ ₯:last_hidden_state[batch, seq, 1024]. - νλ§: λͺ¨λΈμ last_hidden_state λ§ μΆλ ₯ β νλ§/μ κ·ν(masked mean pool + L2)λ μλΉμΈ‘μμ μν. 1024-dim.
- ν ν¬λμ΄μ : XLM-R SentencePiece (
sentencepiece.bpe.modelλλ΄ β xlm-roberta-large/BGE-M3 μ λμΌ, sha256cfc8146aβ¦).
μμ± λ°©λ² (μ¬ν)
python export_embedder_onnx.py --model nlpai-lab/KURE-v1 --out ./kure-v1 --dtype int8
torch.onnx.export(legacy, eager attention) β quantize_dynamic(QInt8). μ μΆλ ₯ μ΄λ¦μ μλΉ C++ μλ² λ κ³μ½ κ³ μ .
λΌμ΄μ μ€
μλ³Έ nlpai-lab/KURE-v1 (MIT) μ νμ. MIT, μμ μμ νκΈ°.
- Downloads last month
- 9
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support