Your next ML (Cloud) Infrastructure for your Code

Which ML Framework is best for the new CLOUD infrastructure (independent if NVIDIA H100 or GOOGLE TPUs)? The future of Machine Learning Accelerators (NVIDIA Tensor Core H100 GPU and Google's TPU Pod v4) w/ ML compiler = XLA.
Plus JAX and TensorFlow3 for new, optimized ML Cloud computing in 2023.
There could be a winner if you want pure speed and auto-cloud-parallelism over 1000s of TPU-chips v4 for you advanced ML models.
#nvidia
#h100
#tpu
#xla
#cloudcomputing

[@johnkost2514]

Excellent walk-thru. Thank-you.

[@code4AI]

Glad you enjoyed it!

[@user-wr4yl7tx3w]

but isn't PyTorch incorporating a lot of the features of JAX?

[@code4AI]

developer are not stupid. but it is a different thing if you try to rebuild your car at 100 km/h (pytorch2) to integrate new compiler, or you build your racing car (JAX) - that is a compiler (smile) - from the very beginning.

[@RajeshSingh-hx2sc]

@@code4AI So learning pytorch wont be any useful if I have to use above infrastructure?