For running CNN model using tensorflow dies for just 30 images for training. This way this device will not be helpful at all for trying out new stuff. Is there a way to overcome this???
The memory for GPU access is limited to the 4GB.
You may try to convert the model into TensorRT to see if can save memory.
Is there any way to increase ram on dev kit??
Sorry that the physical RAM cannot be increased since it is embedded to the Jetson board.
You may try to add some swap space but this kind of memory only can be accessed via CPU.
By the way, Jetson is an edge device and it is not suitable for training.
It’s more recommended to use a desktop GPU or clouds for deep learning training jobs.
I have a similar issue. I have a google browser with a number of tabs and working with making database uodates to an aws cloud. In working with a large number of images being added to the database with details I find I am having to reboot the Jetson a number of times a day because it seems to run low on memory to the point I can’t issue another command as it just hangs for a very long time and I end up having to reboot the Jetson rather than sit for ages waiting for the system to react. I noticed you inidcated that 4Mb is the max RAM for this system and unable to increase it. Is this kind of problem a regular issue or is this model too small for this sort of desktop work? Is there a more powerful model as an option and what are the memory sizes?
Also, in the long term we will be using this type of model to use cameras to read moving images and uploading them to a cloud. Is this model suffice for this sort of thing or will it be a similar issue to the db work described in the previous message?
This should be Tell Loud on Internet: “edge device”