Falcon 180B is the latest addition to TII’s Falcon model family, representing a significant advancement in generative AI.
Key Features:
- It’s a scaled-up version of Falcon 40B, boasting innovations like multiquery attention for enhanced scalability.
- Trained on a massive 3.5 trillion tokens, utilizing up to 4096 GPUs and approximately 7,000,000 GPU hours.
- It’s 2.5 times larger than Llama 2 and benefits from four times more computational resources.
- Training data consists primarily of web content (~85%) and a curated mix of conversations, technical papers, and code (~3%).
- Fine-tuned for versatility on chat and instruction datasets from various conversational sources.
Commercial Use: Commercial use of Falcon 180B is allowed but with strict limitations, excluding “hosting use.” It’s crucial to review the licensing terms and seek legal advice if considering commercial deployment.
Falcon 180B is poised to make significant contributions to generative AI, setting new benchmarks in natural language understanding and generation.
We’ll jump into main topic how to deploy a Falcon-180B in Amazon sage maker. If you want to more about Falcon-180B model and how to load and use model in consumer hardware. please checkout previous blog.