AuroraAI Research has released Aurora1.0-150M, a new 150 million parameter language model. The model's performance is comparable to GPT-2 Small, with benchmark scores including 62.24% on PIQA and 32.20% on Hellaswag. It was trained on 7 billion tokens using an RTX Pro 6000, and an inference script is available on Hugging Face. AI
IMPACT Provides a new, smaller model option for researchers and developers working with limited resources.
RANK_REASON Release of a new, small-scale language model from a research group. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →