跳到正文
Mira Murati· @miramurati · X·· 9 天前AI 评分58
AI 导读

Thinking Machines 在 Connectionism 上详解了 Inkling 2 Small 的端到端 4-bit 训练方法,在 NVIDIA GB300 上从预训练第一步到最终 RL rollout 全程使用原生 4-bit 前向。Mira Murati 表示这是整套技术栈多项改进之一,将预训练时间从数周缩短到 68 小时。引用内容指出,用更少比特训练速度更快,但量化误差可能干扰学习。

正文 · 原文

Today on Connectionism: how we trained Inkling 2 Small with a native 4-bit forward pass on @nvidia GB300s, from the first pretraining step through the final RL rollout, one of the many improvements across the stack that took pretraining from weeks to 68 hours.

引用Thinking Machines@thinkymachines
Training with fewer bits is faster, but quantization error can interfere with learning. Today on Connectionism, we detail how we trained Inkling 2 Small with a native 4-bit forward pass from the first step of pretraining through the final RL rollout. https://thinkingmachines.ai/blog/end-to-end-4-bit-training
在 X 查看被引用的帖子

来源:Mira Murati · x.com