Audit Logs and dataset download in web app, weighted training for Reinforcement Fine-Tuning

Audit Logs in Web App

You can now view and search audit logs directly from the Fireworks web app. The new Audit Logs page provides:

• Search and filter logs by status and timeframe
• Detailed view panel for individual log entries
• Easy navigation from the console sidebar under Account settings

See the Audit Logs documentation for more information.

Dataset Download

You can now download datasets directly from the Fireworks web app. The new download functionality allows you to:

• Download individual files from a dataset
• Download all files at once with "Download All"
• Access downloads from the Datasets table in the dashboard

Weighted Training for Reinforcement Fine-Tuning

Reinforcement Fine-Tuning now supports per-example weighting, giving you more control over which samples have greater influence during training. This feature mirrors the weighted training functionality already available in Supervised Fine-Tuning.

See the Weighted Training documentation for details on the weight field format.

✨ New Models

• KAT Coder is now available in the Model Library

[accordion] Bug Fixes & Minor Improvements

• Console Navigation: Redesigned sidebar with organized groups (CREATE, EXPLORE, MANAGE) for easier navigation (Web App)
• Fine-Tuning Progress Display: Training progress (progress/epoch) now displays inline in the job title while Supervised Fine-Tuning jobs are running (Web App)
• Reinforcement Fine-Tuning Evaluator Selection: Evaluators that are still building are now disabled in the selector with a tooltip and status badge (Web App)
• Loss Chart Smoothing: Large datasets (>1000 metrics) now show EMA smoothing by default for improved visibility (Web App)
• Checkpoint Restore: Fixed base model resolution when promoting checkpoints from fine-tuning jobs (Web App)
• Deployment Usage Charts: Fixed usage graph display on deployment details page (Web App)
• Evaluation Job Share Links: Fixed incorrect output dataset share links and improved deep-link behavior for evaluation jobs (Web App)
• Embedding Model Deployments: Embedding models can now be deployed directly from the UI (Web App)
• Dataset Download State: Download option is now disabled for datasets that are still uploading (Web App)
• GPU Hints: Removed invalid GPU hints from region selector in deployment form (Web App)
• PEFT Model Shapes: Fixed deployment shape lookup for PEFT Addon and Live Merge models (Web App)
• LoRA Validation: Improved error messages when LoRA checkpoints are missing the required language_model. prefix, with actionable conversion instructions (API)
• Reinforcement Fine-Tuning Timeout: Extended maximum job timeout from 4 to 7 days for longer training runs (API)
• Training Job Cancellation: Added ability to cancel Supervised Fine-Tuning, DPO, and Reinforcement Fine-Tuning jobs via API (API)
• Resource Errors: Improved error messages for capacity-related issues during training and deployment (API)