AdamS: Momentum Itself Can Be A Normalizer for LLM Pretraining and Post-training
This work addresses optimization bottlenecks for researchers and practitioners training large language models, offering a more efficient and easy-to-adopt alternative to Adam, though it is incremental as it builds upon existing optimizer frameworks.
The authors tackled the problem of inefficient optimization in large language model training by introducing AdamS, a new optimizer that eliminates second-moment estimates, reducing memory and compute to match SGD with momentum while achieving superior performance in pre-training and post-training tasks, such as on GPT-2 and Llama2 models up to 13B parameters.
We introduce AdamS, a simple yet effective alternative to Adam for large language model (LLM) pretraining and post-training. By leveraging a novel denominator, i.e., the root of weighted sum of squares of the momentum and the current gradient, AdamS eliminates the need for second-moment estimates. Hence, AdamS is efficient, matching the memory and compute footprint of SGD with momentum while delivering superior optimization performance. Moreover, AdamS is easy to adopt: it can directly inherit hyperparameters of AdamW, and is entirely model-agnostic, integrating seamlessly into existing pipelines without modifications to optimizer APIs or architectures. The motivation behind AdamS stems from the observed $(L_0, L_1)$ smoothness properties in transformer objectives, where local smoothness is governed by gradient magnitudes that can be further approximated by momentum magnitudes. We establish rigorous theoretical convergence guarantees and provide practical guidelines for hyperparameter selection. Empirically, AdamS demonstrates strong performance in various tasks, including pre-training runs on GPT-2 and Llama2 (up to 13B parameters) and reinforcement learning in post-training regimes. With its efficiency, simplicity, and theoretical grounding, AdamS stands as a compelling alternative to existing optimizers.