As model sizes increase, ML practitioners may struggle with training costs and time. To make model training faster, cheaper, and with higher accuracy, MosaicML has released Composer. Composer, an OSS library implemented in PyTorch, enables users to implement more than 24 methods to accelerate training loops with a few lines of code through its Functional API. Alternatively, users can apply a built-in Trainer that will automatically implement best practices to achieve optimal performance. Composter also allows users to evaluate and apply combinations of speed-up methods including by handling collisions between methods and ensuring the proper order of execution.