code wiki / lowrank
topic: lowrank
7 modules sharing the lowrank name family (derived from the tree's prefix discipline).
| module | description | lines | funcs |
|---|---|---|---|
| nx_lowrank_attn.nx | M3: low-rank KV through a REAL softmax-attention forward (end-to-end quality proof). | 402 | 20 |
| nx_lowrank_kv_budget.nx | the HARDWARE PAYOFF of data-driven low-rank KV (nx_lowrank_rank_plan) on the | 93 | 7 |
| nx_lowrank_kv_gate.nx | MEASURED Pareto gate: low-rank KV-cache compression via Frequent Directions. | 226 | 10 |
| nx_lowrank_kv_real.nx | M4c-1: is REAL Qwen2.5-0.5B KV actually low-rank? (the empirical foundation of the | 366 | 18 |
| nx_lowrank_rank_plan.nx | DATA-DRIVEN per-tensor/per-layer KV rank sizing. Bridges M2 ("the optimal | 96 | 5 |
| nx_lowrank_train.nx | M2: does a LOW-RANK bottleneck IMPROVE quality (not just preserve it)? The MLA/ | 273 | 13 |
| nx_lowrank_weight_spectrum.nx | the FORWARD-FREE VRAM lever: are the real model's WEIGHT matrices low-rank? | 185 | 14 |