I try to pre-allocate a numpy array, and then assign the value by iteration Buffer = np.zeros((1024, 120, 90, 3), dtype=np.uint8) for i in range(8192): print(f"{i} Memory Usage: {psutil.Process().memory_info().rss / 1e6:.2f} MB") state = np.random.randint(0, 2, (120, 90, 3), dtype=np.uint8) Buffer[i%1024][:] = state print(f"{i} Memory Usage: {psutil.Process().memory_info().rss / 1e6:.2f} MB") And during the loop, the memory usage of process will suddenly increase by 2.1MB in the certain iterations (exists in both first 1024 iterations and later), this cause the process out-of-memory. I have tried adding del state gc.collect() in the end of each iteration, or use Buffer[i%1024] = state for assignment, but both of them didn't work. Can anyone tell me what's happened, and how to fix it? Continue reading...
