Its not the pressure, but the fact water blocks radio waves. Submarines can’t get GPS signals or transmit their location. Submarines have to rely on other means such as IMUs to track their own location.
They have locator beacons size of a thermos (that ping powerfully using particular frequencies) which they fit on flights etc... i believe some can be received a couple miles+ down in good conditions like are seeing now.
But clearly not everything sinks, or the sub would never be able to surface. If it's just a matter of engineering work, that's kind of literally their job.
These things come with a lot of limitations, like not being able to work with 4D and larger tensors, plus the 8 bit quantization. Getting models to run on them is a real pain.
Docker is the only way to do ML. One thing I didn’t think to do before for a while was make separate docker files for each project so all of the deps are installed automatically and that has since helped tremendously.
Wrote this simple one to create gzipped tar backups and send them over ssh.
A lot faster than rsync if you just need to back up a whole folder on a schedule.
It requires pigz, which is a parallel gzip implementation.
Variables:
$HOSTNAME - the computer hostname
$TOBACKUPDIR - the local directory you want backed up
$N_CORES - the number of cores you want to use for compression
$REMOTEUSER - the ssh user login on the remote server
$REMOTEHOST - the remote server's IP
$BACKUPDIR - where you want the file to be backed up to
#!/bin/bash
bfile=`date +%F`.$HOSTNAME.tar.gz
/usr/bin/tar cvpf - \
# You can exclude local directories here with
# --exclude="dir" \
$TOBACKUPDIR | pigz -p $N_CORES | \
ssh $REMOTEUSER@$REMOTEHOST "cat - > /$BACKUPDIR/$bfile"
