Clarifai Release 9.2

Community of Clarifa

Added single sign-on (SSO) sign-up option

You can now use the SSO feature to sign in to the Clarifai platform with your Google account.

Released v2 form for creating deeply trained models

You can now use the new shape to create your deep rendered models. Forms have been updated for a modern look and feel.


  • Tweaked resource titles to improve user-friendliness and structure consistency. Headers for various resources, such as pages for models, workflows, applications, and datasets, are now generated properly and consistently.
  • Improved the design of the pending checkout page. When a new user registers for an account, they are redirected to a pending confirmation page that prompts them to check their email. address before accessing our platform. We have improved the design of the page.
  • Added infinite scrolling to display concepts instead of displaying all concepts at once. The Concepts tab, which lists the concepts available in each model, is now bookmarked. This allows you to easily view the concepts in each model, especially for models with many concepts.
  • Added pagination to the collaboration page. The list of collaborations is now tabbed.
  • Added message to input-manager for unassigned datasets filter option. Before the tooltip Unassigned The datasets filter option was empty. It now has a message that appears when the cursor is placed over it.
  • Renamed some UI fields to create deep built models. We’ve renamed and rearranged some UI fields to create deep built-in models to make them easier to understand.
  • Made improvements to the model viewer video interface. When you upload a video to the model viewer after clicking Try your own image or video button, you’ll get a new and updated interface for video controls, including an option to view video in full screen.
  • Added option to switch between multiple GPUs when training deep models. If you have more than one GPU on your rig, you can now switch between them when training deep models. That way you can take advantage of both GPUs and get faster results.
  • Optimized the user interface on to be better

Bug fixes

  • Fixed an issue where some app names appeared instead of app IDs in the Use in Workflow dropdown. It Use in a workflow a pop-up window appears if you want to use a model in a workflow. Previously, the dropdown that allows you to select a destination application for your workflow could fill in application names instead of application IDs. The issue is now resolved and there are only app IDs to choose from.
  • Fixed an issue where searching for any string was too fast. Previously, search results could appear after each character entered in the search box. Unnecessary search results could not disappear, making it difficult to go back and do a new search after searching for something. Currently, the search is only performed after the user finishes typing the search term.
  • Fixed an issue where a blank section appeared on the model viewer page during forecasting. Previously, an empty box would appear on the prediction load after 0.5 seconds. The model viewer page now loads properly.
  • Fixed an issue where paging a page overwrites the original search results. Pagination no longer rewrites the initial results after performing a search.
  • Fixed an issue where text from one column overflowed into another on the model list page. Text on the model listing page no longer spans multiple columns.
  • Fixed an issue where the user received no feedback after adding a new concept tag in the Input-Viewer. Previously, the user would not receive any notification after successfully adding a new concept to an entry in the Explorer viewport, only until he tried to add it a second time, in which case he would receive a notification that it had a duplicate ID. .
  • Fixed an issue where deep train configuration was reset to default. Previously, when creating a new model of a deep trained visual detector, the image size parameter for the template configuration kept being reset to the default value of 512.


Added ability to overwrite model output_info: in the workflow graph

Users now get more specific error codes that tell them why a particular request failed. Codes reflect errors related to certain parameters, which make debugging problems easier and faster.

  • You can now override the outputs from given nodes in your workflow by passing WorkflowNode.OutputInfoOverride parameter in the definition of each api.WorkflowNode. This is a change ModelVersion you specified and tells us to use the modified one ModelVersion when you use the workflow to infer.
  • If provided, any output_info will dominate the chart output_info from stored model_version during workflow forecasting. This way, if you don’t have the model version, you can change the output settings to suit your specific use cases.
  • This change affects PostWorkflow: endpoint (for adding workflows to applications), PatchWorkflow: endpoint (for editing workflows) and PostWorkflow ResultsRequest: endpoint (for predicting with workflows).

Source link