You no longer need to leave Python to write high-performance hardware kernels. Learn how to use Pallas in Keras to author custom ops that lower to Mosaic for TPUs or Triton for GPUs:

→ View original post on X — @fchollet,