|
ember.cpp 0.1.0
C++ inference engine for Llama-family and paperkiln models (CPU, CUDA, WebAssembly)
|
2-bit K-quantized block structure More...
#include <quantization.h>

Public Attributes | |
| uint16_t | d |
| uint16_t | dmin |
| uint8_t | scales [GGML_QK_K/16] |
| uint8_t | qs [GGML_QK_K/4] |
2-bit K-quantized block structure
Stores weights quantized to 2 bits with block-wise scaling. Provides maximum compression at the cost of precision.
Definition at line 102 of file quantization.h.
| uint16_t block_q2_K::d |
Block scale
Definition at line 103 of file quantization.h.
| uint16_t block_q2_K::dmin |
Block minimum value
Definition at line 104 of file quantization.h.
Quantized values
Definition at line 106 of file quantization.h.
Sub-block scales
Definition at line 105 of file quantization.h.