ROCm/FastFlowLM v0.1.7

🧬 FastFlowLM v0.1.7 — Full Qwen3 Family Support

on GitHub

FastFlowLM v0.1.7 expands model coverage by adding support for the complete Qwen3 family, enabling flexible deployment across different model sizes.

✨ What's New

✅ Full Qwen3 Model Family Support

• Added support for the following Qwen3 variants:
  • Qwen3:0.6b
  • Qwen3:1.7b
  • Qwen3:4b
  • Qwen3:8b

These additions make FastFlowLM capable of running Qwen3 models from lightweight to large-scale, all optimized for AMD NPUs.

✅ Usage Examples

Interactive CLI:

flm run qwen3:0.6b
flm run qwen3:4b prompt.txt

Server Mode:

flm serve  # Default model still applies unless changed via API

Model Management:

flm list
flm remove qwen3:1.7b

This release continues our commitment to high-performance, NPU-optimized inference for a growing variety of LLM architectures.