Modelpublic

second-state/jina-embeddings-v3-GGUF

sourceHugging Faceupdated 1y agoView on Hugging Face

3likes1.6kdownloads

Model Card

<div style="width: auto; margin-left: auto; margin-right: auto"> <img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </div> <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">

jina-embeddings-v3-Embedding-GGUF

Original Model

jinaai/jina-embeddings-v3

Run with LlamaEdge

—LlamaEdge version: coming soon

—Prompt template

—Prompt type: embedding

—Embedding size: 32, 64, 128, 256, 512, 768, 1024

—Run as LlamaEdge service

bash

  wasmedge --dir .:. --nn-preload default:GGML:AUTO:jina-embeddings-v3-f16.gguf \
    llama-api-server.wasm \
    --prompt-template embedding \
    --ctx-size 768 \
    --model-name jina-embeddings-v3

<!--

Quantized GGUF Models

Name	Quant method	Bits	Size	Use case
jina-embeddings-v3-Q2_K.gguf	Q2_K	2	60.8 MB	smallest, significant quality loss - not recommended for most purposes
jina-embeddings-v3-Q3_K_L.gguf	Q3KL	3	80.6 MB	small, substantial quality loss
jina-embeddings-v3-Q3_K_M.gguf	Q3KM	3	76.2 MB	very small, high quality loss
jina-embeddings-v3-Q3_K_S.gguf	Q3KS	3	68.7 MB	very small, high quality loss
jina-embeddings-v3-Q4_0.gguf	Q4_0	4	83.7 MB	legacy; small, very high quality loss - prefer using Q3KM
jina-embeddings-v3-Q4_K_M.gguf	Q4KM	4	90.0 MB	medium, balanced quality - recommended
jina-embeddings-v3-Q4_K_S.gguf	Q4KS	4	84.0 MB	small, greater quality loss
jina-embeddings-v3-Q5_0.gguf	Q5_0	5	97.9 MB	legacy; medium, balanced quality - prefer using Q4KM
jina-embeddings-v3-Q5_K_M.gguf	Q5KM	5	103 MB	large, very low quality loss - recommended
jina-embeddings-v3-Q5_K_S.gguf	Q5KS	5	97.9 MB	large, low quality loss - recommended
jina-embeddings-v3-Q6_K.gguf	Q6_K	6	113 MB	very large, extremely low quality loss
jina-embeddings-v3-Q8_0.gguf	Q8_0	8	146 MB	very large, extremely low quality loss - not recommended
jina-embeddings-v3-f16.gguf	f16	16	274 MB	very large, extremely low quality loss - not recommended

--> Quantized with llama.cpp b6311