This commit is contained in:
2024-07-12 17:12:17 +08:00
parent 80373c3dc5
commit ba94e917e9
2 changed files with 218 additions and 0 deletions

View File

@@ -0,0 +1,41 @@
services:
ldh-deepspeed-test:
build:
context: .
dockerfile: Dockerfile.ldh
image: ldh/deepspeed:test
shm_size: '32gb'
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
runtime: nvidia
environment:
- NVIDIA_VISIBLE_DEVICES=all
- NVIDIA_DRIVER_CAPABILITIES=compute,utility
stdin_open: true
tty: true
privileged: true
cap_add:
- IPC_LOCK
devices:
- /dev/infiniband:/dev/infiniband
volumes:
- /mnt/local-nvme2:/root/data/local-nvme2
- /mnt/local-nvme:/root/data/local-nvme
# ports:
# - "22242:22242"
# - "5000:5000"
# networks:
# - ldh_overlay_network
network_mode: host
# command: ["/usr/sbin/sshd", "-D"]
# command: ["/usr/sbin/sshd"]
# networks:
# ldh_overlay_network:
# external: true