🤔The Falcon 180B is an enormous model with 180 billion parameters.
🚀It is currently topping the open source LLM leaderboards and rivals proprietary models like GPT-2.
💪The Falcon 180B is a scaled-up version of Falcon 40B, another foundational model.
⏱️It was trained on 3.5 trillion tokens on up to 4096 GPUs simultaneously.
🛠️Although commercially viable, there are restrictive conditions for its use.
How many parameters does the Falcon 180B have?
—The Falcon 180B has 180 billion parameters.
How does the Falcon 180B compare to other models?
—It is currently topping the open source LLM leaderboards and rivals proprietary models like GPT-2.
Is the Falcon 180B an improved version of Falcon 40B?
—Yes, the Falcon 180B is a scaled-up version of Falcon 40B.
How was the Falcon 180B trained?
—It was trained on 3.5 trillion tokens on up to 4096 GPUs simultaneously.
Can the Falcon 180B be used commercially?
—Yes, but under restrictive conditions excluding any hosting use.
00:00Introduction to the Falcon 180B and the question of whether size matters.
01:32Overview of the Falcon 180B's architecture and its relationship to Falcon 40B.
02:52Details about the training process, including the number of tokens and GPUs used.
03:59Explanation of the Falcon 180B's commercial viability and the conditions for its use.
