Pinned Loading

1. Fisher-Adaptive-Binary-Quantization Fisher-Adaptive-Binary-Quantization Public

   FABQ-RC: Fisher-Adaptive Binary Quantization with Residual Codebooks

   Jupyter Notebook 3

2. epic-quant epic-quant Public

   EPIC-Quant: Epi-Stochastic Predictive Fetching & Context-Aware Bit-Shifting for Gemma 4 E4B. CPU-first, real weights, three pillars: attention bit-discarding, PLE sparse caching, KV eviction.

   Python

3. Quantization-Exploration Quantization-Exploration Public

   Experiments in ultra-low bit quantization (≤0.7 bits/weight) for LLM inference, focused on Gemma 4 E2B/E4B

   Python

4. Gemma4-E4B-ple-coded-gguf Gemma4-E4B-ple-coded-gguf Public

   PLE-Coded GGUF: exploiting Gemma E4B's per-layer embeddings as a compression side-channel

   Python