Output-aware recurrent-state quantization with measured GDN model paths, GDN2 operator demos, and reproducible quality, storage, and cost evidence.
-
Updated
Sep 11, 2026 - Python
#
recurrent-state
Here are 2 public repositories matching this topic...
Experiments with mixed INT4/INT8 storage for Qwen3.5 recurrent state.
research transformers pytorch quantization mixed-precision linear-attention cache-compression llm llm-inference qwen qwen3 qwen35 gated-deltanet recurrent-state
-
Updated
Aug 26, 2026 - Python
Add this topic to your repo
To associate your repository with the recurrent-state topic, visit your repo's landing page and select "manage topics."