Why We Started with JAX but Moved to PyTorch

A story of why compute infra is an inherent part of an ML training stack

saved by

• Miguel Villa Floran