ClearML/clearml
1
0
Fork 0
mirror of https://github.com/clearml/clearml synced 2025-05-15 18:05:40 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Update Auto Scaler default values and configuration format

Browse Source
This commit is contained in:
allegroai 2021-02-01 23:44:01 +02:00
parent 12d51dae92
commit 71c45fcdd3
3 changed files with 17 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
clearml/automation/auto_scaler.py
View File

@ -33,7 +33,8 @@ class AutoScaler(object):
class Configuration(object): class Configuration(object):
resource_configurations = attr.ib(default=None) resource_configurations = attr.ib(default=None)
queues = attr.ib(default=None) queues = attr.ib(default=None)
extra_trains_conf = attr.ib(default="") extra_trains_conf = attr.ib(default="") # Backwards compatibility
extra_clearml_conf = attr.ib(default="")
extra_vm_bash_script = attr.ib(default="") extra_vm_bash_script = attr.ib(default="")
def as_dict(self): def as_dict(self):
@ -61,7 +62,7 @@ class AutoScaler(object):
self.cloud_credentials_region = settings.cloud_credentials_region self.cloud_credentials_region = settings.cloud_credentials_region
self.default_docker_image = settings.default_docker_image self.default_docker_image = settings.default_docker_image
self.extra_trains_conf = configuration.extra_trains_conf self.extra_clearml_conf = configuration.extra_clearml_conf or configuration.extra_trains_conf
self.extra_vm_bash_script = configuration.extra_vm_bash_script self.extra_vm_bash_script = configuration.extra_vm_bash_script
self.resource_configurations = configuration.resource_configurations self.resource_configurations = configuration.resource_configurations
self.queues = configuration.queues self.queues = configuration.queues

2
clearml/automation/aws_auto_scaler.py
View File

@ -87,7 +87,7 @@ class AwsAutoScaler(AutoScaler):
queue=queue_name, queue=queue_name,
git_user=self.git_user or "", git_user=self.git_user or "",
git_pass=self.git_pass or "", git_pass=self.git_pass or "",
clearml_conf='\\"'.join(self.extra_trains_conf.split('"')), clearml_conf='\\"'.join(self.extra_clearml_conf.split('"')),
bash_script=self.extra_vm_bash_script, bash_script=self.extra_vm_bash_script,
docker="--docker '{}'".format(self.default_docker_image) docker="--docker '{}'".format(self.default_docker_image)
if self.default_docker_image if self.default_docker_image

24
examples/services/aws-autoscaler/aws_autoscaler.py
View File

@ -1,3 +1,4 @@
import json
from argparse import ArgumentParser from argparse import ArgumentParser
from collections import defaultdict from collections import defaultdict
from itertools import chain from itertools import chain
@ -80,7 +81,8 @@ def main():
# from here on everything is logged automatically # from here on everything is logged automatically
task = Task.init(project_name="DevOps", task_name="AWS Auto-Scaler", task_type=Task.TaskTypes.service) task = Task.init(project_name="DevOps", task_name="AWS Auto-Scaler", task_type=Task.TaskTypes.service)
task.connect(hyper_params) task.connect(hyper_params)
task.connect_configuration(configurations) configurations.update(json.loads(task.get_configuration_object(name="General") or "{}"))
task.set_configuration_object(name="General", config_text=json.dumps(configurations, indent=2))
if args.remote or args.run: if args.remote or args.run:
print("Running AWS auto-scaler as a service\nExecution log {}".format(task.get_output_log_web_page())) print("Running AWS auto-scaler as a service\nExecution log {}".format(task.get_output_log_web_page()))
@ -159,14 +161,14 @@ def run_wizard():
), ),
"ami_id": get_input( "ami_id": get_input(
"the Amazon Machine Image id", "the Amazon Machine Image id",
"['ami-07c95cafbb788face']", "['ami-04c0416d6bd8e4b1f']",
question='Select', question='Select',
default="ami-07c95cafbb788face", default="ami-04c0416d6bd8e4b1f",
), ),
"ebs_device_name": get_input( "ebs_device_name": get_input(
"the Amazon EBS device", "the Amazon EBS device",
"['/dev/xvda']", "['/dev/sda1']",
default="/dev/xvda", default="/dev/sda1",
), ),
"ebs_volume_size": input_int( "ebs_volume_size": input_int(
"the Amazon EBS volume size", "the Amazon EBS volume size",
@ -175,8 +177,8 @@ def run_wizard():
), ),
"ebs_volume_type": get_input( "ebs_volume_type": get_input(
"the Amazon EBS volume type", "the Amazon EBS volume type",
"['gp2']", "['gp3']",
default="gp2", default="gp3",
), ),
"key_name": get_input( "key_name": get_input(
"the Amazon Key Pair name", "the Amazon Key Pair name",
@ -209,10 +211,10 @@ def run_wizard():
) )
print("Entered {} lines of pre-execution bash script".format(num_lines_bash_script)) print("Entered {} lines of pre-execution bash script".format(num_lines_bash_script))
configurations.extra_trains_conf, num_lines_trains_conf = multiline_input( configurations.extra_clearml_conf, num_lines_clearml_conf = multiline_input(
"\nEnter anything you'd like to include in your trains.conf file []" "\nEnter anything you'd like to include in your clearml.conf file []"
) )
print("Entered {} extra lines for trains.conf file".format(num_lines_trains_conf)) print("Entered {} extra lines for clearml.conf file".format(num_lines_clearml_conf))
print("\nDefine the machines budget:") print("\nDefine the machines budget:")
print("-----------------------------") print("-----------------------------")
@ -274,7 +276,7 @@ def run_wizard():
if not input_bool("Do you wish to add another instance type to queue? [y/N]: "): if not input_bool("Do you wish to add another instance type to queue? [y/N]: "):
break break
if not input_bool("\nAdd another queue? [y/N]: "): if not input_bool("\nAdd another queue? [y/N]"):
break break
configurations.queues = dict(queues) configurations.queues = dict(queues)