structure.md

.
├── ...
    ├── layers                      # Low-bit layers
    │   ├── qconv                   # quantized Convolotional Layer 
    │   ├── qembedding              # currently only 1-bit embedding layer supported
    │   ├── ... 
    │   └── qlinear                 
    │       ├── binary (1-bit)
    │       │   ├── cpp             # x86 CPU
    │       │   ├── cuda            # Nvidia GPU
    │       │   └── cutlass         # Nvidia GPU
    │       └── n-bit (2/4/8-bit)
    │           ├── mps             # Apple GPU
    │           ├── cuda            # Nvidia GPU, e.g., weight-only quantized LLMs
    │           └── cutlass         # Nvidia GPU, e.g., quantization aware training for both activation and weight
    └── optim
    │    └── DiodeMix               # dedicated optimizer for low-bit quantized model 
    ├── functions
    └── ...