4 ms·
Not really, for GPU computation, PyTorch, TensorFlow or JAX are the baselines (they are not just for ML stuff). Even for CPU, that's a more relevant comparison.
by albertzeyer 3y ago
Not really, for GPU computation, PyTorch, TensorFlow or JAX are the baselines (they are not just for ML stuff). Even for CPU, that's a more relevant comparison.
- KeplerBoy 3y agoI agree in principle, but my experience is that researchers writing code just default to NumPy.