"llm/vscode:/vscode.git/clone" did not exist on "a00fac4ec8f04b78981e0955c7750e78a79df49e"
  • Daniël de Kok's avatar
    Add support for Marlin 2:4 sparsity (#2102) · f1f98e36
    Daniël de Kok authored
    This change adds support for 2:4 sparsity when using Marlin
    quantization. The 2:4 kernel is used when:
    
    * The quantizer is `marlin`;
    * the quantizer checkpoint format is `marlin_24`.
    
    Fixes #2098.
    f1f98e36
mma.h 7.91 KB