Onboard Qwen3.5 397b fp8 - #5182

Open

Shuwen-Fang wants to merge 12 commits into

pr/fp8-qwen3.5-35b-onboardingfrom

qwen3.5-397b-fp8

Commits on Sep 2, 2026

feat(layers): implement FP8 weight-only dynamic dequantization engine and types
snehalv2002
committed
f74726d
refactor(layers): prune speculative boilerplate and scope strictly to llama 3.1 and qwen 3.5 fp8
snehalv2002
committed
7bddd99

Commits on Sep 4, 2026

feat(layers): support generic weight_block_size configuration for block-scaled quantization
snehalv2002
committed
9f2d431

Commits on Sep 5, 2026

refactor(layers): clarify DenseGeneral scale casework and restore attend_dtype
snehalv2002
committed
f77f5ef

Commits on Sep 8, 2026

Commits on Sep 9, 2026

onboard 397b
Shuwen-Fang
committed
1bb4b03