Release 10.7

Release Date: August 6th, 2024

---

New Feature	Improvement	Bug Fix	Enterprise Only

Llama 3.1

Status	Change	Details
	Introduced the Llama 3.1 template	You can now use the Llama 3.1 training template to fine-tune text-to-text models for classification and generation tasks. We provide the 8B version, which you can use for extended context, instruction-following, and advanced applications. Learn more about it here.

New Published Models

Status	Change	Details
	Published new models (Clarifai-hosted models are the ones we host within our Clarifai Cloud. Wrapped models are those hosted externally, but we deploy them on our platform using their third-party API keys)	Published Qwen1.5-7B-Chat, an open-source, multilingual LLM with 32K token support, excelling in language understanding, alignment with human preferences, and competitive tool-use capabilities. Published Qwen2-7B-Instruct, a state-of-the-art multilingual language model with 7.07 billion parameters, excelling in language understanding, generation, coding, and mathematics, and supporting up to 128,000 tokens. Published Whisper-Large-v3, a Transformer-based speech-to-text model showing 10-20% error reduction compared to Whisper-Large-v2, trained on 1 million hours of weakly labeled audio, and can be used for translation and transcription tasks. Published Llama-3-8b-Instruct-4bit, an instruction-tuned LLM optimized for dialogue use cases. It can outperform many of the available open source chat LLMs on common industry benchmarks. Published Mistral-Nemo-Instruct, a state-of-the-art 12B multilingual LLM with a 128k token context length, optimized for reasoning, code generation, and global applications. Published Phi-3-Mini-4K-Instruct, a 3.8B parameter small language model offering state-of-the-art performance in reasoning and instruction-following tasks. It outperforms larger models with its high-quality data training. Published GPT-4o-mini, an affordable, high-performing small model excelling in text and vision tasks with extensive context support. Published Llama 3.1-8b-Instruct, a multilingual, highly capable LLM optimized for extended context, instruction-following, and advanced applications.

Python SDK

Status	Change	Details
	Added patch operations	Introduced patch operations for input annotations and concepts. Introduced patch operations for apps and datasets.
	Improved the RAG SDK	We enabled the RAG SDK to use environment variables for enhanced security, flexibility, and simplified configuration management.
	Improved the logging experience	Enhanced the logging experience by adding a constant width value to rich logging.
	Fixed an issue with the dataset export feature	We fixed an issue with the feature using a class input session with an authentication header when downloading the dataset export.
	Fixed an issue with Python SDK's visual search functionality	Previously, the Python SDK's visual search functionality returned duplicate results. We fixed the issue, and now the search and ranking results for image searches are returned correctly without duplicates.

Organization Settings and Management​

Status	Change	Details
	Introduced a new Organization User role	This role has access privileges similar to those of an Organization Contributor for all apps and scopes. However, it comes with view-only permissions without create, update, or delete privileges.
	Implemented restrictions on the ability to add new organizations based on the user's current organization count and feature access	If a user has created one organization and does not have access to the multiple organizations feature, the "Add an organization" button is now disabled. We also display an appropriate tooltip to them. If a user has access to the multiple organizations feature but has reached the maximum creation limit of 20 organizations, the "Add an organization" button is disabled. We also display an appropriate tooltip to them.

Modules

Status	Change	Details
	Improved the functionality of the Hyperparamater Sweeps module	You can now use the module to effectively train your model on a range and combinations of hyperparameter values.

Docs Refresh

Status	Change	Details
	Made significant improvements to our documentation site	Upgraded the site to use Docusaurus version 3.4. Other enhancements include aesthetic updates, a more intuitive menu-based navigation, and a new comprehensive API reference guide.

Models

Status	Change	Details
	Enabled deletion of associated model assets when removing a model annotation	Now, when deleting a model annotation, the associated model assets are also marked as deleted.

Workflows

Status	Change	Details
	Improved the functionality of the Face workflow	You can now use the Face workflow to effectively generate face landmarks and perform face visual searches within your applications.

Community

Status	Change	Details
	Fixed issues with Python and Node.js SDK code snippets	If you click the “Use Model” button on an individual model’s page, the “Call by API / Use in a Workflow” modal appears. You can then integrate the displayed code snippets in various programming languages into your own use case. Previously, the code snippets for Python and Node.js SDKs for image-to-text models incorrectly outputted concepts instead of the expected text. We fixed the issue to ensure the output is now correctly provided as text.
	Added support for non-ASCII characters	Previously, non-ASCII characters were entirely filtered out from the UI when creating concepts. We fixed this issue, and you can now use non-ASCII characters across all components.
	Improved the display of concept relations	Concept relations are now shown next to their respective concept names, providing clearer and more immediate context.
	Fixed an issue with delay and redirection during login	Previously, a modal login form would appear when attempting to log in at https://clarifai.com/explore. However, a second login form would also appear, causing confusion with the presence of two forms. We fixed the issue, ensuring a smoother and more intuitive login experience.
	Fixed an issue with "View inputs" link on smaller screens	Previously, the "View inputs" link on the App Overview page could exceed its boundary on smaller screens, particularly if the number of inputs in the app exceeded 1,000. We fixed this issue to ensure the link now properly adjusts to screen size, maintaining its alignment and readability.

Auto-Annotation

Status	Change	Details
	Fixed some bugs related to the auto-annotation feature	Fixed an issue where some inputs were not labeled during the auto-annotation process. Fixed an issue where auto-generated concept IDs were incorrectly displayed instead of concept names.

Labeling Tasks

Status	Change	Details
	Fixed an issue with the labeling tasks table	Previously, the date column was not appearing in the table for listing labeling tasks. The date column now correctly displays in the table, providing the necessary date information for each labeling task.
	Fixed an issue where an unnecessary modal popup appeared on the tasks review screen	Previously, when you created a new task, labeled it with appropriate concepts, and clicked the Review button to navigate to the review screen, a modal stating "No task assignments" could immediately appear. We fixed this issue, and the modal no longer appears, ensuring a smoother review process.
	Fixed an issue with annotation display on the labeling tasks list	Previously, annotations were displayed with annotation IDs instead of names on the labeling tasks list. We fixed this issue, and annotations are now correctly displayed with their names.
	Fixed an issue with creating a dataset version from task page	Previously, when creating a dataset version from the task page for an auto-annotation task that had been reviewed, the version included labels rejected during the review process. We fixed this issue. Now, when creating a dataset version from the task page, the current state of inputs is accurately considered, and any rejected labels or inputs are excluded from the version.

Input-Viewer

Status	Change	Details
	Fixed an issue with bounding box labeling on the Input-Viewer screen	Previously, during labeling with a bounding box on the single Input-Viewer screen, the bounding box could sometimes start from the wrong position, requiring users to adjust it manually. We fixed this issue, ensuring a smoother labeling experience.