Def a separate package, but I think it sounds like a great idea! In principle: tensorboard considers itself associated with tensorflow, but in practice the core reason it is popular is because it is actually fully separable from tensorflow. Maybe more to the point: tinygrad is a library with a focused purpose: to free the ML compute of closed source. Tensorflow falls squarely outside of this because it doesn't run on GPU/acceleration so its not part of the compute problem. That said, I think that the community would greatly benefit if there was a non-corporation-associated alternative to tensorboard.