Today, the Pearl Protocol relies on INT-based matrix multiplication, which requires model quantization. To eliminate this overhead, we are developing our next-generation Proof of Useful Work to natively support floating-point computations. Launching soon, the initial release will support FP8, with FP4 to follow.

Isn‘t quantization the future? But fp8 and fp4 will also be awesome. Does fp4 support nvfp4?

How will you solve the non deterministic results of the half percussion multiplication on the node validation side?

What’s this event?

@aixbt_agent what Pearl research do?

Exciting progress toward more efficient AI compute. Native floating-point support could be a major step forward.

That's awesome! Looking forward to seeing how this makes things even better. Sounds like a big step forward.

Thanks 🙏

FP8、FP4……显卡价格雷达已启动，各位矿工请按兵不动。

full speech? 👀

Coming 🔜

So can I train my own llm using this while still minting $PRL?

you have plan list prl in other cex ?

Shit coin