ClearML FAQ | Hi All, Idk Whether This Is The Right Channel To Ask This Question My Team Are Using Clearml As Part Of Our Development Process. We Wish To Use Clearml-Serve As Well To Deploy Our Services To K8S; However, As Clearml-Serve Is Not Mature Enough, We’Ll Wai

Answered

Hi All, Idk Whether This Is The Right Channel To Ask This Question My Team Are Using Clearml As Part Of Our Development Process. We Wish To Use Clearml-Serve As Well To Deploy Our Services To K8S; However, As Clearml-Serve Is Not Mature Enough, We’Ll Wai

Hi all,
Idk whether this is the right channel to ask this question

My team are using ClearML as part of our development process.
We wish to use ClearML-serve as well to deploy our services to k8s; however, as ClearML-serve is not mature enough, we’ll wait for 1.0 version (more backends are supported) and we are finding an alternative approach to use in the meantime.

What we are thinking is how to build a bridge to “link” ClearML-server (act as a model registry) and the deployment environment (k8s for example). Then, using DevOps to automatically build service, get chosen models from ClearML server and deploy the service to the staging and production environments.
Any ideas, guys?

  				
Posted 
	2 years ago

					More
				  		
  Report
		
					GrittyKangaroo27
				
					0
					 × 1

Votes Newest

Answers

Hi GrittyKangaroo27
Maybe check the TriggerScheduler , and have a function trigger something on k8s every time you "publish" a model?
https://github.com/allegroai/clearml/blob/master/examples/scheduler/trigger_example.py

  				
Posted 
	2 years ago

					More
				  		
  Report
		
					AgitatedDove14
				
					0
					 × 1

Write your answer

1K Views

1 Answer

2 years ago

one year ago

Tags

Similar posts

The Clearml Enterprise Plan Supports K8S Servicing Multiple Clearml Queues, As Well As Providing A Pod Template For Each Queue For Describing The Resources For Each Pod To Use

New To Clearml, I Primarily Use Rye To Handle My Package Management. For Some Reason, Our Training Task Is Capturing All Of Our Depenendencies As Declared In Our

Hey! We'Ve Set An Hyperparameter Optimization Task On One Of Our Experiments, And It Seems As Though The Optimization Itself Is Running - Yet We Don'T See The Iterations Table In The 'Plots' Section In The Ui. Providing Some Screenshots To How We'Ve Set I

Hello Everyone, We Are Currently Looking Into Clearml To Possibly Integrate Into Our Ml Workflow. We Are Wondering How The Integration Is With Orchestrators Such As Dagster. Can We Still Define Tasks That Can Be Queued Through The Ui And Hadled By Agents?

Hello Team, We Are Currently Utilizing A Self-Hosted Clearml For Various Projects. Our Aim Is To Implement Role-Based Access Control Across Projects. As I Understand, This Feature Is Presently Available Only On The Enterprise Plan. Is It Possible To Enab

Hi Everyone. I Am Trying To Migrate To Clearml And Need To Have My Old Training Logs Available In Clearml As Well. Unfortunately It Seems I Can'T Simply "Import" My Old Tensorboard Logs Into Clearml To Have It All In One Place. Does Any One Have A Suggest

Hi, We Are Using Clearml For Our Experiment Tracking But Now Investigating Using The Pipeline Functionality As Well For Scheduling. We Also Want To Be Able To Trigger A Pipeline Run When There Is New Data In An External Database. Is This Possible? From Wh

Hi, We'Ve Been Trying To Run The Same Experiments As We Did At Previous Times But Have Been Getting Configuration Errors After Credentials Errors Over And Over. I'Ve Went Ahead And Replaced All Of Our Existing Credentials With New Ones, Went Through The

Hello, Are There Any Plans To Add Support For Pdm Package Manager? It'S What We Use Since Poetry Dependcy Solver Is Quite Slow And It Would Be Neat To Have Direct Support In Clearml-Agents As Well. Thanks!

We Do Want To Have Control Which Files Are Logged In The Model Registry. There Is Such Option In Task.Init(), Auto_Connect_Frameworks=False Or Injecting A Dict() To It.. But Our Tasks Are Created As Being Part Of A Pipeline With Add_Function_Step(). So We