Hugging Face's logo Hugging Face
  • Models
  • Datasets
  • Spaces
  • Buckets
  • Docs
  • Enterprise
  • Pricing
    • Website
      • Tasks
      • HuggingChat
      • Collections
      • Languages
      • Organizations
    • Community
      • Blog
      • Posts
      • Daily Papers
      • Hardware
      • Learn
      • Discord
      • Forum
      • GitHub
    • Solutions
      • Team & Enterprise
      • Hugging Face PRO
      • Enterprise Support
      • Inference Providers
      • Inference Endpoints
      • Storage Buckets

  • Log In
  • Sign Up

HelloSun
/
sdcpp_zimage_opt

GGUF
stable-diffusion.cpp
stable-diffusion
z-image
z-image-turbo
cpu
benchmark
amx
avx512
operator-fusion
Model card Files Files and versions
xet
Community
sdcpp_zimage_opt
6.76 MB
Ctrl+K
Ctrl+K
  • 1 contributor
History: 9 commits
HelloSun's picture
HelloSun
stage3: also record F32 conv GEMM negative result
5ec9c4d about 23 hours ago
  • patches
    VAE stage 3: vectorize direct-conv im2col (VAE decode ~4.7s -> ~3.9s, pixel-identical) about 23 hours ago
  • results
    Stage 1 (VAE decode): llamafile build, conv-direct tests, additive 8-step rows; code + results 1 day ago
  • stages
    stage3: also record F32 conv GEMM negative result about 23 hours ago
  • tools
    Pipeline stage P2: explain and fix refiner Q8_0 slowdown, V-pack hoist + 128x128 flash-attn tiles about 23 hours ago
  • .gitattributes
    3.48 kB
    Stage 1 (VAE decode): llamafile build, conv-direct tests, additive 8-step rows; code + results 1 day ago
  • README.md
    12.1 kB
    VAE stage 3: vectorize direct-conv im2col (VAE decode ~4.7s -> ~3.9s, pixel-identical) about 23 hours ago
  • SmartTest.sh
    11.2 kB
    Pipeline stage P2: explain and fix refiner Q8_0 slowdown, V-pack hoist + 128x128 flash-attn tiles about 23 hours ago
  • build.sh
    10.7 kB
    VAE stage 3: vectorize direct-conv im2col (VAE decode ~4.7s -> ~3.9s, pixel-identical) about 23 hours ago