Falcon3-10B-Base

<div align="center"> <img src="https://huggingface.co/datasets/tiiuae/documentation-images/resolve/main/general/falco3-logo.png" alt="drawing" width="500"/> </div>

Falcon3-10B-Base

Falcon3 family of Open Foundation Models is a set of pretrained and instruct LLMs ranging from 1B to 10B parameters.

This repository contains the Falcon3-10B-Base. It achieves state-of-the-art results (at the time of release) on reasoning, language understanding, instruction following, code and mathematics tasks. Falcon3-10B-Base supports 4 languages (English, French, Spanish, Portuguese) and a context length of up to 32K.

⚠️ This is a raw, pretrained model, which should be further finetuned using SFT, RLHF, continued pretraining, etc. for most use cases.

Model Details

• Architecture
  • Transformer-based causal decoder-only architecture
  • 40 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
• Depth up-scaled from Falcon3-7B-Base with continual pretraining on 2 Teratokens of datasets comprising of web, code, STEM, high quality and mutlilingual data using 1024 H100 GPU chips
• Supports EN, FR, ES, PT
• Developed by Technology Innovation Institute
• License: TII Falcon-LLM License 2.0
• Model Release Date: December 2024

Getting started

<details> <summary> Click to expand </summary>

import torch
from transformers import pipeline

pipe = pipeline(
    "text-generation", 
    model="tiiuae/Falcon3-10B-Base", 
    torch_dtype=torch.bfloat16, 
    device_map="auto"
)
response = pipe("Question: How many hours in one day? Answer: ")
print(response[0]['generated_text'])

</details> <br>

Benchmarks

We report in the following table our internal pipeline benchmarks.

• We use lm-evaluation harness.
• We report raw scores.
• We use same batch-size across all models.

<table border="1" style="width: 100%; text-align: center; border-collapse: collapse;"> <colgroup> <col style="width: 10%;"> <col style="width: 10%;"> <col style="width: 7%;"> <col style="width: 7%;"> <col style="width: 7%;"> <col style="background-color: rgba(80, 15, 213, 0.5); width: 7%;"> </colgroup> <thead> <tr> <th>Category</th> <th>Benchmark</th> <th>Gemma2-9B</th> <th>Yi1.5-9B</th> <th>Mistral-Nemo-Base-2407 (12B)</th> <th>Falcon3-10B-Base</th> </tr> </thead> <tbody> <tr> <td rowspan="3">General</td> <td>MMLU (5-shot)</td> <td>70.8</td> <td>69.6</td> <td>68.8</td> <td><b>73.1</b></td> </tr> <tr> <td>MMLU-PRO (5-shot)</td> <td>41.4</td> <td>39.3</td> <td>34.7</td> <td><b>42.5</b></td> </tr> <tr> <td>IFEval</td> <td>21.3</td> <td>29.1</td> <td>16.1</td> <td><b>36.4</b></td> </tr> <tr> <td rowspan="2">Math</td> <td>GSM8K (5-shot)</td> <td>69.1</td> <td>63.8</td> <td>55.3</td> <td><b>81.4</b></td> </tr> <tr> <td>MATH Lvl-5 (4-shot)</td> <td>10.5</td> <td>9.2</td> <td>4.9</td> <td><b>22.9</b></td> </tr> <tr> <td rowspan="4">Reasoning</td> <td>Arc Challenge (25-shot)</td> <td>67.5</td> <td>61.7</td> <td>64.4</td> <td><b>66.8</b></td> </tr> <tr> <td>GPQA (0-shot)</td> <td>33.4</td> <td><b>36.6</b></td> <td>28.8</td> <td>34.1</td> </tr> <tr> <td>MUSR (0-shot)</td> <td><b>45.3</b></td> <td>43.3</td> <td>39.2</td> <td>44.2</td> </tr> <tr> <td>BBH (3-shot)</td> <td>54.3</td> <td>51.3</td> <td>50.2</td> <td><b>59.7</b></td> </tr> <tr> <td rowspan="4">CommonSense Understanding</td> <td>PIQA (0-shot)</td> <td><b>83.0</b></td> <td>80.5</td> <td>82.1</td> <td>79.4</td> </tr> <tr> <td>SciQ (0-shot)</td> <td><b>97.1</b></td> <td>95.2</td> <td>95.2</td> <td>93.5</td> </tr> <tr> <td>Winogrande (0-shot)</td> <td><b>74.2</b></td> <td>72.7</td> <td>73.2</td> <td>73.6</td> </tr> <tr> <td>OpenbookQA (0-shot)</td> <td><b>47.2</b></td> <td>45.2</td> <td><b>47.2</b></td> <td>45.0</td> </tr> </tbody> </table>

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}
}

Open LLM Leaderboard Evaluation Results

Detailed results can be found here