Hi There, Is There A Way To Save A Model Simply To The Fileserver Such That The Model Url Will Point There And Not To A Local Disk (I Am Running In Docker Containers And Local Path Is Not Always Available).

Answered

Hi there, is there a way to save a model simply to the fileserver such that the MODEL URL will point there and not to a local disk (I am running in docker containers and local path is not always available).

  				
Posted 
	7 months ago

					More
				  		
  Report
		
					MagnificentBear85
				
					0
					 × 1

Votes Newest

Answers 3

Setting sdk.development.default_output_uri to the same URL as api.files_server seems to do the trick.

  				
Posted 
	7 months ago

					More
				  		
  Report
		
					MagnificentBear85
				
					0
					 × 1

Hi @<1577468611524562944:profile|MagnificentBear85> , you can also use Task.init(... , output_uri=True) in the code as well

  				
Posted 
	7 months ago

					More
				  		
  Report
		
					CostlyOstrich36
				
					0

Ah, thanks!

  				
Posted 
	7 months ago

					More
				  		
  Report
		
					MagnificentBear85
				
					0
					 × 1

Write your answer

603 Views

3 Answers

7 months ago

Hi Again! Are There Way Of Uploading Your Model Architecture To Clearml, And Not The Weights. Would Like To Easily Compare Different Experiments With Slightly Different Architectures And See The Difference In How Data Flows Through The Model.

Hello Everyone, I Have A Quick Question, I Am Using Clearml For An Ml Experiment Tracking Project. As Is, Clearml Is Saving A Version Of My Model After Each Epoch. Is There A Way For Clearml To Simply Save The Model Once Training Is Done And To Ignore The

I Wonder If There Is Any Way To Sync Data On Cloud Storage (E.G S3 Or Minio) Using Clearml-Data Sync Command? Because I Don’T See Any Place In The Documentation Mention It, We Can Upload From Local Files But Not The Files On Cloud

Hi, I Have An Artifact Which Is A Path To The Outputs Of The Model That Was Trained And Its Used In Another Task Later On But As A String, But When I Use Upload Artifact It Uploads The Whole Folder And Not Just As A String, Is There Are Way To Just Upload

I Seem To Be Missing Something ... I'Ve Only Got One Task Running To Train A Segmentation Model On My Local Machine, And In A Few Days It'S Hit Over 1.15M Api Calls. It Looks Like It'S Sending Every Single Console Output ... Are There Settings To Control

Hi All! I Want To Setup A Tool That I Can Use To Analyse Clearml Logs (Just User And Experiment Logs, I Am Not Doing Any Model Serving Yet). Is There An Api Endpoint That I Can Fetch Such Metrics? Metrics Like Number Of Active Users, Number Of Experiments

Is It Necessary To Serve Keras Model Using Triton Engine? I'M Trying To Serve An Endpoint, And Trying To Debug, But The Error Given Not Helping Much. Is There A Flag I Can Pass To See More Logs?

Is There A Way To Serve A Model Via The Sdk Or Rest Api? I Want To Programmatically Serve The Model After Finishing Training It Via The Pipeline. Or Is It A Bad Practice To Do So, Hence Why It'S Not Really Exposed?

Hi! I Am Trying To Update Existing Models That Were Already Created By Some Model Registration Task. I Can Not Find In The Documentation If It Is Possible To Update The Model'S Tag Programmatically? There Is A Way To Update The Model'S Tag From The Web Pa