Some Like It Small: Czech Semantic Embedding Models for Industry Applications
Paper • 2311.13921 • Published
How to use from
llama.cppInstall from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Veol-CZ/czech-semantic-embeddings-c:F16# Run inference directly in the terminal:
llama cli -hf Veol-CZ/czech-semantic-embeddings-c:F16Use pre-built binary
# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Veol-CZ/czech-semantic-embeddings-c:F16# Run inference directly in the terminal:
./llama-cli -hf Veol-CZ/czech-semantic-embeddings-c:F16Build from source code
git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Veol-CZ/czech-semantic-embeddings-c:F16# Run inference directly in the terminal:
./build/bin/llama-cli -hf Veol-CZ/czech-semantic-embeddings-c:F16Use Docker
docker model run hf.co/Veol-CZ/czech-semantic-embeddings-c:F16Quick Links
Czech Semantic Embeddings (GGUF)
GGUF conversions of Czech semantic embedding models from Seznam, prepared for llama.cpp embedding inference.
Included Models
Seznam__retromae-small-cs.f16.ggufSeznam__retromae-small-cs.q8_0.ggufSeznam__dist-mpnet-paracrawl-cs-en.f16.ggufSeznam__dist-mpnet-paracrawl-cs-en.q8_0.ggufSeznam__simcse-retromae-small-cs.f16.ggufSeznam__simcse-retromae-small-cs.q8_0.ggufSeznam__simcse-dist-mpnet-paracrawl-cs-en.f16.ggufSeznam__simcse-dist-mpnet-paracrawl-cs-en.q8_0.gguf
Upstream Sources
- https://github.com/seznam/czech-semantic-embedding-models
- https://huggingface.co/Seznam/retromae-small-cs
- https://huggingface.co/Seznam/dist-mpnet-paracrawl-cs-en
- https://huggingface.co/Seznam/simcse-retromae-small-cs
- https://huggingface.co/Seznam/simcse-dist-mpnet-paracrawl-cs-en
Citation
If you use these models, please cite the original Seznam paper:
- Bednar et al., 2024, Some Like It Small: Czech Semantic Embedding Models for Industry Applications
- https://arxiv.org/abs/2311.13921
@inproceedings{bednavr2024some,
title={Some Like It Small: Czech Semantic Embedding Models for Industry Applications},
author={Bedn{\'a}{\v{r}}, Ji{\v{r}}{\'\i} and N{\'a}plava, Jakub and Baran{\v{c}}{\'\i}kov{\'a}, Petra and Lisick{\`y}, Ond{\v{r}}ej},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={38},
number={21},
pages={22734--22742},
year={2024}
}
Usage (llama.cpp)
Run embedding server:
llama-server -m Seznam__retromae-small-cs.q8_0.gguf --embedding --pooling cls
OpenAI-compatible embeddings endpoint:
curl http://127.0.0.1:8080/v1/embeddings \
-H "Content-Type: application/json" \
-d '{"input":["Ahoj svete","Semantic search test"]}'
File Integrity
SHA256 checksums are in checksums.txt.
License
This repo contains converted checkpoints from upstream models. Respect original model licenses and terms:
retromae-small-cs: CC-BY-4.0dist-mpnet-paracrawl-cs-en: CC-BY-4.0simcse-retromae-small-cs: CC-BY-4.0simcse-dist-mpnet-paracrawl-cs-en: CC-BY-4.0
Attribution to Seznam and original model cards is required.
- Downloads last month
- 33
Hardware compatibility
Log In to add your hardware
8-bit
16-bit
Install (macOS, Linux)
# Start a local OpenAI-compatible server with a web UI: llama serve -hf Veol-CZ/czech-semantic-embeddings-c:F16# Run inference directly in the terminal: llama cli -hf Veol-CZ/czech-semantic-embeddings-c:F16