You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Program prints the following output (on the first and second nodes respectively):
['multinode.py', '192.168.6.2:9000', '0']
hello 0
['multinode.py', '192.168.6.2:9000', '1']
hello 1
No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)
1
So, the first node didn't print the number of local devices, the second one couldn't even find them. However, when I run jax scripts which use all gpus on the first/second node, everything works as expected.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I try to run simple multi-host program on two nodes (each node has 4 gpus) which are connected via ethernet:
On the first node I use this command
NCCL_SOCKET_IFNAME=enp36s0f1 python3 multinode.py <ip>:9000 0
On the second node I use this command:
NCCL_SOCKET_IFNAME=enp36s0f1 python3 multinode.py <ip>:9000 1
Program prints the following output (on the first and second nodes respectively):
So, the first node didn't print the number of local devices, the second one couldn't even find them. However, when I run jax scripts which use all gpus on the first/second node, everything works as expected.
Beta Was this translation helpful? Give feedback.
All reactions