Halong Embedding Gguf

🚀 hiieu/halong_embedding in GGUF format

This project provides the hiieu/halong_embedding model in GGUF format. It offers quantization scripts and usage examples for sentence similarity tasks.

📦 Installation

The installation process mainly involves downloading necessary files and setting up the environment. Here are the steps:

REL=b3827  # can change to a later release
wget https://github.com/ggerganov/llama.cpp/releases/download/$REL/llama-$REL-bin-ubuntu-x64.zip --content-disposition --continue &> /dev/null
wget https://github.com/ggerganov/llama.cpp/archive/refs/tags/$REL.zip                           --content-disposition --continue &> /dev/null
unzip -q llama-$REL-bin-ubuntu-x64.zip
unzip -q llama.cpp-$REL.zip
mv llama.cpp-$REL/* .
rm -r llama.cpp-$REL/ llama-$REL-bin-ubuntu-x64.zip llama.cpp-$REL.zip
pip install -q -r requirements.txt

rm -rf models/tmp/
git clone --depth=1 --single-branch https://huggingface.co/hiieu/halong_embedding models/tmp
huggingface-cli download intfloat/multilingual-e5-base sentencepiece.bpe.model --local-dir models/tmp
python convert_hf_to_gguf.py models/tmp/ --outfile model-f32.gguf --outtype f32

build/bin/llama-quantize model-f32.gguf model-f16.gguf    f16    2> /dev/null
build/bin/llama-quantize model-f32.gguf model-bf16.gguf   bf16   2> /dev/null
build/bin/llama-quantize model-f32.gguf model-q8_0.gguf   q8_0   2> /dev/null
build/bin/llama-quantize model-f32.gguf model-q6_k.gguf   q6_k   2> /dev/null
build/bin/llama-quantize model-f32.gguf model-q5_k_m.gguf q5_k_m 2> /dev/null
build/bin/llama-quantize model-f32.gguf model-q5_k_s.gguf q5_k_s 2> /dev/null
build/bin/llama-quantize model-f32.gguf model-q4_k_m.gguf q4_k_m 2> /dev/null
build/bin/llama-quantize model-f32.gguf model-q4_k_s.gguf q4_k_s 2> /dev/null

rm -rf models/yolo/
mkdir -p models/yolo
mv model-*.gguf models/yolo/
touch models/yolo/README.md
huggingface-cli upload halong-embedding-gguf models/yolo .

💻 Usage Examples

Basic Usage

build/bin/llama-embedding -m model-q5_k_m.gguf -p "Cô ấy cười nói suốt cả ngày" --embd-output-format array 2> /dev/null

Advanced Usage

build/bin/llama-server --embedding -c 512 -m model-q5_k_m.gguf

📄 License

This project is licensed under the CC BY 4.0 license.

Additional Information