how can I use my IB 100g card to training with multi node? about colossalai HOT 3 CLOSED

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

Describe the feature

root@MZ32-00:/ColossalAI/examples/language/llama2# ifconfig

eno1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.1.11 netmask 255.255.255.0 broadcast 192.168.1.255
inet6 2409:8a50:3020:5d70:e29b:8cc1:fa33:cbd8 prefixlen 64 scopeid 0x0
inet6 fe80::5259:de5e:e208:a742 prefixlen 64 scopeid 0x20
inet6 2409:8a50:3020:5d70:bb44:a058:849b:d594 prefixlen 64 scopeid 0x0
ether 18:c0:4d:80:51:56 txqueuelen 1000 (Ethernet)
RX packets 2701416 bytes 543187328 (543.1 MB)
RX errors 1 dropped 4880 overruns 0 frame 1
TX packets 20864211 bytes 30107024017 (30.1 GB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
device memory 0x94420000-9443ffff

ibp65s0: flags=4099<UP,BROADCAST,MULTICAST> mtu 4092
inet 11.4.3.177 netmask 255.255.0.0 broadcast 11.4.255.255
unspec 00-00-00-68-FE-80-00-00-00-00-00-00-00-00-00-00 txqueuelen 256 (UNSPEC)
RX packets 0 bytes 0 (0.0 B)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 0 bytes 0 (0.0 B)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

++++++++++++++++++++++++++++++++++

amd01@MZ32-01:~$ ifconfig

eno2: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.1.12 netmask 255.255.255.0 broadcast 192.168.1.255
inet6 fe80::97d6:6dda:5bee:7a41 prefixlen 64 scopeid 0x20
inet6 2409:8a50:3020:5d70:33a8:55ce:93da:d5a9 prefixlen 64 scopeid 0x0
inet6 2409:8a50:3020:5d70:bcb5:c325:76ac:ac60 prefixlen 64 scopeid 0x0
ether d8:5e:d3:12:29:ea txqueuelen 1000 (Ethernet)
RX packets 95709 bytes 109708513 (109.7 MB)
RX errors 0 dropped 1485 overruns 0 frame 0
TX packets 173482 bytes 134568240 (134.5 MB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
device memory 0x94400000-9441ffff

ibp65s0: flags=4099<UP,BROADCAST,MULTICAST> mtu 4092
inet 11.4.3.175 netmask 255.255.0.0 broadcast 11.4.255.255
unspec 00-00-00-68-FE-80-00-00-00-00-00-00-00-00-00-00 txqueuelen 256 (UNSPEC)
RX packets 0 bytes 0 (0.0 B)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 0 bytes 0 (0.0 B)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
++++++++++++++++++++++++++++++

now, I can run

root@MZ32-00:/ColossalAI/examples/language/llama2# colossalai run --nproc_per_node 2 --host 192.168.1.11,192.168.1.12 --master_port=29507 --master_addr 192.168.1.11 pretrain.py

well, but failed on:
root@MZ32-00:/ColossalAI/examples/language/llama2# colossalai run --nproc_per_node 2 --host 11.4.3.177,11.4.3.175 --master_port=29590 --master_addr 11.4.3.177 pretrain.py
WARNING:torch.distributed.run:

Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.

Error: failed to run torchrun --nproc_per_node=2 --nnodes=2 --node_rank=1 --master_addr=11.4.3.177 --master_port=29590 pretrain.py on 11.4.3.175, is localhost: False, exception: [Errno None] Unable to connect to port 22 on 11.4.3.175

from colossalai.

Hi, have you solved this problem? I'm facing a similar issue.

from colossalai.

Yes, I did

But I am not sure which step I take.

The problemn is the config of ubuntu, not ColossalAI

from colossalai.