Low Memory Warning" warning from the system!
NVIDIA's message was that it was a low-cost version, even if it was a forum!
Even with the increased computing speed, the problem was the swap memory: the Jetson Orin Nano has 8GB of memory, which was not enough to hold the training data, so it had to read and write to the microSD swap memory, which took time in areas other than AI inference.
It is beyond the scope of the hobbyist Sunday programmer, although it is necessary to devise programs to fit into 8GB of memory, such as by dividing data and processing it, or by other methods that aim to both lower memory usage and speed up the process. Compared to the old jetson-nano, it is not as user-friendly for the price considering the upper limit of personal spending range, and it seems to be a product that is dragging down GPU-CUDA.
At this price, it is better to buy a PC, and raspi has released version 5/8G at 1/5 the price.
NVIDIA's target may be the interdisciplinary research field, but I wish they would be more gentle.
I will leave you with a program that I have tested the operation of.
... thi site can not paste the program...sorry.
import pkg_resources for _lib in pkg_resources.working_set: print(_lib.project_name, _lib.version)
from tensorflow.python.client import device_lib device_lib.list_local_devices()
import tensorflow as tf; print( tf.__version__ )
# https://www.tensorflow.org/tutorials/quickstart/beginner
import tensorflow as tf print("TensorFlow version:", tf.__version__)
mnist = tf.keras.datasets.mnist (x_train, y_train), (x_test, y_test) = mnist.load_data() x_train, x_test = x_train / 255.0, x_test / 255.0
. . . . . . . . . .
