qingy2024/LLaMa_3.2_3B_Catalysts

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:3.2BQuant:BF16Ctx Length:32kPublished:Oct 19, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

qingy2024/LLaMa_3.2_3B_Catalysts is a 3 billion parameter Llama-3.2-based instruction-tuned model developed by qingy2019. It was fine-tuned using Unsloth and Huggingface's TRL library, enabling 2x faster training. This model is evaluated on the Open LLM Leaderboard, showing performance across various benchmarks including IFEval, BBH, and MMLU-PRO.

Loading preview...

Model Overview

qingy2024/LLaMa_3.2_3B_Catalysts is a 3 billion parameter language model developed by qingy2019. It is fine-tuned from the unsloth/Llama-3.2-3B-Instruct base model, leveraging the Unsloth library for accelerated training (2x faster) in conjunction with Huggingface's TRL library.

Key Capabilities & Performance

This model's performance is evaluated on the Open LLM Leaderboard, providing insights into its general capabilities. Notable benchmark results include:

  • Avg. Score: 19.63
  • IFEval (0-Shot): 49.92
  • BBH (3-Shot): 21.35
  • MATH Lvl 5 (4-Shot): 11.10
  • GPQA (0-shot): 5.15
  • MuSR (0-shot): 7.95
  • MMLU-PRO (5-shot): 22.31

Training Details

The model was trained with Unsloth, a framework designed to speed up the fine-tuning process for large language models, making it an efficient choice for developers looking for Llama-3.2 based models with optimized training. The full evaluation details are available on the Hugging Face Open LLM Leaderboard dataset.