ROCm/FastFlowLM v0.1.5

🔒 FastFlowLM v0.1.5 — Template Engine, NPU Locking & Usability Boosts

on GitHub

FastFlowLM v0.1.5 enhances stability, improves API integration, and introduces convenient new features for both developers and users.

✨ What's New

1. Chat Template Engine via Google Minja

• Integrated Google Minja to handle chat templates.
• This improves compatibility with standardized prompt formats and makes template management easier.

2. Optimized Serve Mode Logging

• Cleaner, more structured logging in serve mode for easier monitoring and debugging.

3. Desktop Shortcut for Server Launch

• A desktop shortcut is now provided to quickly launch flm serve without using the terminal.

4. NPU Access Lock

• Introduced an NPU access lock mechanism.
• Prevents simultaneous conflicting NPU requests by blocking access when the NPU is already in use.

5. Updated CLI Commands

• Improved and expanded CLI commands for a better command-line experience.

✅ Usage Recap

Interactive CLI:

flm run llama3.2
flm run llama3.2 prompt.txt

Server Mode (via CLI or Desktop Shortcut):

flm serve

Model Management:

flm help
flm list
flm remove llama3.2:1b

This release improves both backend stability and frontend usability, while laying the foundation for broader chat template support.