Playground category tabs, Contributor and Inference roles, and stop and resume for fine-tuning jobs

Playground Categories

The Playground now features category tabs (LLM, Image, TTS, STT) in the header for easier switching between model types. The playground automatically detects the appropriate category based on the selected model and provides smart defaults for each category.

User Roles: Contributor and Inference

New user roles provide more granular access control for team collaboration:

• Contributor: Read and write access to resources without administrative privileges
• Inference: Read-only access with the ability to run inference on deployments

Assign these roles when inviting team members to provide appropriate access levels.

Fine-Tuning Improvements

Fine-tuning workflows have been enhanced with several new capabilities:

• Stop and Resume Jobs: Stop running fine-tuning jobs and resume them later from where they left off. Available for Supervised Fine-Tuning and Reinforcement Fine-Tuning jobs.
• Clone Jobs: Quickly create new fine-tuning jobs based on existing job configurations using the Clone action.
• Download Output Datasets: Download output datasets from Reinforcement Fine-Tuning jobs, including individual files or bulk download as a ZIP archive.
• Download Rollout Logs: Download rollout logs from Reinforcement Fine-Tuning jobs for offline analysis.

✨ New Models

• Gemma 3 12B Instruct is now available in the Model Library
• Gemma 3 4B Instruct is now available in the Model Library
• Qwen3 Omni 30B A3B Instruct is now available in the Model Library

📚 Documentation Updates

• Deployment Shapes API: Added List Deployment Shape Versions and Get Deployment Shape endpoints for querying available deployment shapes
• Evaluator APIs: Added Create Evaluator, Update Evaluator, and helper endpoints for evaluator source code, build logs, and upload validation
• Fine-Tuning APIs: Added Resume DPO Job, Resume Reinforcement Fine-Tuning Step, Execute Reinforcement Fine-Tuning Step, and Get Evaluation Job Log Endpoint
• SDK Examples: Added Python SDK example links for direct routing and supervised fine-tuning workflows

[accordion] Bug Fixes & Minor Improvements

• Deployment Progress Display: Deployment details page now shows live deployment progress with replica status (pending, downloading, initializing, ready) and error banners (Web App)
• Multi-LoRA Display: Deployment details page now shows all deployed models with expandable sections, not just the default model (Web App)
• Audio Usage Charts: Audio usage metrics are now displayed in a dedicated Voice tab on the Usage page with filtering support (Web App)
• Deployment Shape Search: Improved deployment shape discovery for models where exact base model matches aren't found, using parameter count bucketing (Web App)
• Vision Model Auto-Detection: Vision-language models (Qwen VL, LLaVA, Phi-3 Vision, etc.) now automatically have image input support enabled when uploaded (API)
• Dataset Loading UX: Output dataset tables now stream results with a progress indicator for faster perceived loading (Web App)
• File Size Limit: Dataset uploads now enforce a clear file size limit with improved error messaging (Web App)
• Number Input Fields: Improved number input validation across forms (Web App)
• Combobox Responsiveness: Improved combobox dropdown height on smaller screens (Web App)
• Evaluator Editor Scroll: Prevented accidental page navigation when scrolling inside the evaluator code editor (Web App)
• Evaluator Save Dialog: Fixed overflow issues in the save evaluator dialog (Web App)
• Evaluator Selector Labels: Fixed label rendering in async evaluator select components (Web App)
• Deploy Button Validation: Deploy button is now disabled when the model is not ready (Web App)
• Model Metadata: Fixed missing model metadata display on deployment pages (Web App)
• Invoice Display: Invoice list now shows "paid" status for contract payments (Web App)
• Color Palette Update: Updated UI color palette for improved visual consistency (Web App)