Modalities, a PyTorch-native Framework For Large-scale LLM Training and Research
This addresses the issue for LLM researchers who face high compute costs and limited tooling in conducting large-scale experiments, though it is incremental as it builds on existing parallelization strategies and framework concepts.
The paper tackles the problem of inefficient and poorly supported large-scale ablation studies in LLM training by introducing Modalities, a PyTorch-native framework that integrates parallelization for efficient pretraining and systematic ablations at trillion-token and billion-parameter scale, with modular design for improved reproducibility and extensibility.
Today's LLM (pre-) training and research workflows typically allocate a significant amount of compute to large-scale ablation studies. Despite the substantial compute costs of these ablations, existing open-source frameworks provide limited tooling for these experiments, often forcing researchers to write their own wrappers and scripts. We propose Modalities, an end-to-end PyTorch-native framework that integrates data-driven LLM research with large-scale model training from two angles. Firstly, by integrating state-of-the-art parallelization strategies, it enables both efficient pretraining and systematic ablations at trillion-token and billion-parameter scale. Secondly, Modalities adopts modular design with declarative, self-contained configuration, enabling reproducibility and extensibility levels that are difficult to achieve out-of-the-box with existing LLM training frameworks.