ClearML/clearml-agent
1
0
Fork 0
mirror of https://github.com/clearml/clearml-agent synced 2025-01-31 09:06:52 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Fix nvidia docker support on some linux distros (SUSE)

Browse Source
This commit is contained in:
allegroai 2020-11-25 16:44:37 +02:00
parent de332b9e6b
commit 64e198a57a
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
trains_agent/commands/worker.py
View File

@ -542,6 +542,10 @@ class Worker(ServiceCommandSection):
'--full-monitoring' if self._services_mode else '--disable-monitoring', '--full-monitoring' if self._services_mode else '--disable-monitoring',
'--standalone-mode' if self._standalone_mode else '', '--standalone-mode' if self._standalone_mode else '',
task_id) task_id)
# send the actual used command line to the backend
self.send_logs(task_id=task_id, lines=['Executing: {}\n'.format(full_docker_cmd)], level="INFO")
cmd = Argv(*full_docker_cmd) cmd = Argv(*full_docker_cmd)
print('Running Docker:\n{}\n'.format(str(cmd))) print('Running Docker:\n{}\n'.format(str(cmd)))
else: else:
@ -2296,7 +2300,7 @@ class Worker(ServiceCommandSection):
if os.environ.get('TRAINS_DOCKER_SKIP_GPUS_FLAG', None): if os.environ.get('TRAINS_DOCKER_SKIP_GPUS_FLAG', None):
dockers_nvidia_visible_devices = gpu_devices dockers_nvidia_visible_devices = gpu_devices
else: else:
base_cmd += ['--gpus', 'device='+gpu_devices, ] base_cmd += ['--gpus', '\"device={}\"'.format(gpu_devices), ]
# We are using --gpu, so we should not pass NVIDIA_VISIBLE_DEVICES, I think. # We are using --gpu, so we should not pass NVIDIA_VISIBLE_DEVICES, I think.
# base_cmd += ['-e', 'NVIDIA_VISIBLE_DEVICES=' + gpu_devices, ] # base_cmd += ['-e', 'NVIDIA_VISIBLE_DEVICES=' + gpu_devices, ]
elif gpu_devices.strip() == 'none': elif gpu_devices.strip() == 'none':