Falcon3-7B-Instruct achieves state-of-the-art results (at release's time) on reasoning, language understanding, instruction following, code and mathematics tasks. Falcon3-7B-Instruct supports 4 languages (English, French, Spanish, Portuguese) and a context length of up to 32K.

This repository contains the GGUFs instruction-tuned 7B Falcon3 model.

Model Details

Architecture
- Transformer-based causal decoder-only architecture
- 28 decoder blocks
- Grouped Query Attention (GQA) for faster inference: 12 query heads and 4 key-value heads
- Wider head dimension: 256
- High RoPE value to support long context understanding: 1000042
- Uses SwiGLU and RMSNorm
- 32K context length
- 131K vocab size
Pretrained on 14 Teratokens of datasets comprising of web, code, STEM, high quality and mutlilingual data using 1024 H100 GPU chips
Posttrained on 1.2 million samples of STEM, conversational, code, safety and function call data
Supports EN, FR, ES, PT
Developed by Technology Innovation Institute
License: TII Falcon-LLM License 2.0
Model Release Date: December 2024
Quantization: q2_K, q3_K_M, q4_0, q4_K_M, q5_0, q5_K_M, q6_K, q8_0

Getting started

1. Download GGUF models from hugging face

First, download the model from Hugging Face. You can use the huggingface_hub library or download it manually:

bash

pip  install  huggingface_hub
huggingface-cli  download  {model_name}

This will download the model to your current directory. Make sure to replace {model_name} with the actual username and model name from your Hugging Face repository.

2. Install llama.cpp

You have several options for installing llama.cpp:

1. Build from source:

This gives you the most flexibility and control. Follow the instructions in the llama.cpp repository to build from source:

bash

git  clone  https://github.com/ggerganov/llama.cpp
cd llama.cpp
cmake -B build
cmake --build build --config Release

For more information about how to build llama.cpp from source please refere to llama.cpp documentation on how to build from source: llama.cpp build from source.

2. Download pre-built binaries:

If you prefer a quicker setup, you can download pre-built binaries for your operating system. Check the llama.cpp repository for available binaries.

3. Use Docker:

For a more contained environment, you can use the official llama.cpp Docker image. Refer to the llama.cpp documentation for instructions on how to use the Docker image.

For detailed instructions and more information, please check the llama.cpp documentation on docker: llama.cpp docker.

3. Start playing with your model

Run simple text completion

llama-cli -m {path-to-gguf-model} -p "I believe the meaning of life is" -n 128

Run in conversation mode

llama-cli -m {path-to-gguf-model} -p "You are a helpful assistant" -cnv -co

Useful links

View our release blogpost.
Feel free to join our discord server if you have any questions or to interact with our researchers and developers.

Technical Report

Coming soon....

Citation

If the Falcon3 family of models were helpful to your work, feel free to give us a cite.

@misc{Falcon3,
    title = {The Falcon 3 Family of Open Models},
    url = {https://huggingface.co/blog/falcon3},
    author = {Falcon-LLM Team},
    month = {December},
    year = {2024}
}

tiiuae/Falcon3-7B-Instruct-GGUF

Falcon3-7B-Instruct-GGUF