Data-driven mixed-precision native TurboQuant checkpoint of Qwen/Qwen3.6-35B-A3B. Extends -TQ-apex2 by skipping the shared-expert down-projection — the tensor family our per-tensor kurtosis scan flagged as 100 % FP16-bucket (κ mean 16.3, max 67, 41 of 41 tensors above threshold). apex2 raised those to TQ4; apex3 leaves them uncompressed. This is the endpoint of PAT-0349 Kurtosis-Aware Automatic Bit-Width Selection for this model: every tensor family whose measured κ puts it in the FP16 bucket is now actually stored as FP16. 1. sharedexpert.downproj → FP16 (41 tensors, ~600 MB). The largest remaining κ-bucket mismatch in apex2 — TQ4’s 16 centroids can’t represent a tensor whose distribution has 67× Gaussian tail mass without discarding outliers. Leaving them uncompressed is cheap on disk (~600 MB on 18 GB) and eliminates the main residual quality leak. 2. Routed downproj → TQ3 (41 tensors, −40 MB). apex / apex2 caught these under the generic downproj substring in SENSITIVEPATTERNS and bumped them to TQ4 — but routed down_proj is κ mean 3.4, which PAT-0349 puts in the TQ3 bucket. TQ4 was wasted bits. Net disk-size delta: apex2 → apex3 is +600 MB − 40 MB ≈ +560 MB. Expected total…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: varjosoft
Теги: qwen3_5_moe, image-text-to-text, turboquant, tq3, tq4, higgs, compressed, quantized
Лайков: 4 | Загрузок: 206
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
