Aurora1.0-150M Releases!

The first generation of the Aurora1.0-150M model has been released, demonstrating performance similar to GPT2-Small. Benchmarks show scores such as PIQA: 62.24%, Hellaswag: 32.20%, Arc-Easy: 44.91%, Arc-Challenge: 25.00%, Arithmark 3.0: 33.90%, and CapitalBench: 36.55%. The model was trained on 7B tokens using an RTX Pro 6000, and an example inference script is available in its Huggingface repository.