Knowledge distillation is an approach for transferring the knowledge of a much larger teacher model to a smaller student model, ideally yielding a compact, easily deployable student with comparable accuracy to the teacher. Knowledge distillation has gained popularity in pretraining settings, but there are fewer resources available for performing knowledge distillation during supervised fine-tuning…
]]>This post was originally published August 21, 2024 but has been revised with current data. Recently, NVIDIA and Mistral AI unveiled Mistral NeMo 12B, a leading state-of-the-art large language model (LLM). Mistral NeMo 12B consistently outperforms similarly sized models on a wide range of benchmarks. We announced Mistral-NeMo-Minitron 8B, one of the most advanced open-access models in…
]]>Large language models (LLM) are now a dominant force in natural language processing and understanding, thanks to their effectiveness and versatility. LLMs such as Llama 3.1 405B and NVIDIA Nemotron-4 340B excel in many challenging tasks, including coding, reasoning, and math. They are, however, resource-intensive to deploy. As such, there is another trend in the industry to develop small language…
]]>Training with larger batches is a straightforward way to scale training of deep neural networks to larger numbers of accelerators and reduce the training time. However, as the batch size increases, numerical instability can appear in the training process. The purpose of this post is to provide an overview of one class of solutions to this problem: layer-wise adaptive optimizers, such as LARS, LARC…
]]>