If you want CUDA up front go write PyTorch. No one is stopping you. Modular’s goal was to leverage MLIR first and bring GPUs in later. They’re barely a year old

If you want CUDA up front go write PyTorch. No one is stopping you. Modular’s goal was to leverage MLIR first and bring GPUs in later. They’re barely a year old company.