Hyperparameter (HP) tuning in deep learning is an expensive process, prohibitively so for neural networks (NNs) with billions of parameters. We show that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes. This leads to a new HP tuning paradigm we call muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tun...| papers.cool
本文解读一下我们最新的技术报告《Muon is Scalable for LLM Training》，里边分享了我们之前在《Muon优化器赏析：从向量到矩阵的本质跨越》介绍过的Muon优化器的一...| kexue.fm
Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon a...| papers.cool
随着LLM时代的到来，学术界对于优化器的研究热情似乎有所减退。这主要是因为目前主流的AdamW已经能够满足大多数需求，而如果对优化器“大动干戈”，那么需要巨大的验证成本。因此，当前优化器的变化，...| kexue.fm
随着算力的飞速进步，有越多越多的场景希望能够实现“算力换时间”，即通过堆砌算力来缩短模型训练时间。理想情况下，我们希望投入$n$倍的算力，那么达到同样效果的时间则缩短为$1/n$，此时总的算力成...| kexue.fm
事实上，除了写博客内容，在这几年里，笔者是花了相当一部分时间来做科学空间的“表面功夫”，为此还专门学了一点php、css和js。虽然不敢说精益求精，但总体来说网站的浏览体验应该比前几年要好得多。...| kexue.fm