Jetson Inference hacks Jan 18, 2022 Releasing the memory occupied by tensorflow from numba import cuda device = cuda.get_current_device() device.reset() Please enable JavaScript to view the comments powered by Disqus.