kyujinpy/PlatYi-34B-Llama-Q
PlatYi-34B-Llama-Q is a 34 billion parameter auto-regressive language model developed by Kyujin Han (kyujinpy), based on the Yi-34B transformer architecture. Fine-tuned using Q-LoRA on the Open-Platypus dataset, it demonstrates strong performance across various benchmarks, achieving an average score of 71.13 on the Open LLM Leaderboard. This model is optimized for general text generation and reasoning tasks, showing notable improvements in GSM8K over its base model.
Loading preview...
PlatYi-34B-Llama-Q: An Enhanced Yi-34B Model
PlatYi-34B-Llama-Q is a 34 billion parameter auto-regressive language model developed by Kyujin Han (kyujinpy). It is built upon the robust Yi-34B transformer architecture, specifically leveraging the chargoddard/Yi-34B-Llama base model. This model has been fine-tuned using Q-LoRA with a lora_r value of 64, utilizing the garage-bAInd/Open-Platypus dataset.
Key Capabilities & Performance
PlatYi-34B-Llama-Q excels in general language understanding and generation tasks. Its fine-tuning process has led to improved benchmark performance compared to its base models. On the Open LLM Leaderboard, it achieves an impressive average score of 71.13.
Key benchmark results include:
- ARC (25-Shot): 65.70
- HellaSwag (10-Shot): 85.22
- MMLU (5-Shot): 78.78
- TruthfulQA (0-shot): 53.64
- Winogrande (5-shot): 83.03
- GSM8K (5-shot): 60.42
Notably, it shows a significant improvement in GSM8K (mathematical reasoning) compared to the PlatYi-34B-Llama and Yi-34B models, making it a strong candidate for tasks requiring logical and arithmetic capabilities.
Good For
- General text generation and understanding applications.
- Tasks requiring improved reasoning, particularly in mathematical contexts (e.g., problem-solving).
- Developers looking for a fine-tuned Yi-34B variant with enhanced benchmark performance.