Parallelization

Parallelization is a core feature of DataDreamer that allows for performant workflows. Advanced users may want to implement parallelization in different ways, depending on their use case. This section will cover the different ways parallelization can be implemented in DataDreamer, including device parallelization (multi-GPU inference and training).

• Running Steps in Parallel
• Running Models on Multiple GPUs
• Training Models on Multiple GPUs