v5.6.2
📦 transformersView on GitHub →
🐛 2 fixes
Summary
This patch release fixes critical issues related to Qwen MoE models using FP8 and improves kernel configuration handling.
🐛 Bug Fixes
- Fixed an issue where Qwen 3.5 and 3.6 MoE (text-only) models failed when used with FP8.
- Fixed configuration reading and error handling for kernels.