Add Q2_0 quantization: type definition and CPU backend (#24448)

This commit is contained in:
Pasha Khosravi
2026-07-07 12:05:47 -07:00
committed by GitHub
parent c198af4dc2
commit bec4772f6a
17 changed files with 262 additions and 3 deletions
+7
View File
@@ -17,6 +17,7 @@
#define ggml_vec_dot_mxfp4_q8_0_generic ggml_vec_dot_mxfp4_q8_0
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q1_0_q8_0_generic ggml_vec_dot_q1_0_q8_0
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
#define ggml_vec_dot_tq1_0_q8_K_generic ggml_vec_dot_tq1_0_q8_K
#define ggml_vec_dot_tq2_0_q8_K_generic ggml_vec_dot_tq2_0_q8_K
#define ggml_vec_dot_q2_K_q8_K_generic ggml_vec_dot_q2_K_q8_K
@@ -82,6 +83,7 @@
#define ggml_gemm_q2_K_8x8_q8_K_generic ggml_gemm_q2_K_8x8_q8_K
#elif defined(__x86_64__) || defined(__i386__) || defined(_M_IX86) || defined(_M_X64)
// quants.c
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
// repack.cpp
#define ggml_quantize_mat_q8_0_4x4_generic ggml_quantize_mat_q8_0_4x4
#define ggml_quantize_mat_q8_K_4x4_generic ggml_quantize_mat_q8_K_4x4
@@ -113,6 +115,7 @@
#define quantize_row_q8_K_generic quantize_row_q8_K
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q1_0_q8_0_generic ggml_vec_dot_q1_0_q8_0
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
#define ggml_vec_dot_tq1_0_q8_K_generic ggml_vec_dot_tq1_0_q8_K
#define ggml_vec_dot_tq2_0_q8_K_generic ggml_vec_dot_tq2_0_q8_K
#define ggml_vec_dot_iq1_m_q8_K_generic ggml_vec_dot_iq1_m_q8_K
@@ -162,6 +165,7 @@
#define ggml_vec_dot_mxfp4_q8_0_generic ggml_vec_dot_mxfp4_q8_0
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q1_0_q8_0_generic ggml_vec_dot_q1_0_q8_0
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
// repack.cpp
#define ggml_quantize_mat_q8_0_4x4_generic ggml_quantize_mat_q8_0_4x4
#define ggml_quantize_mat_q8_0_4x8_generic ggml_quantize_mat_q8_0_4x8
@@ -202,6 +206,7 @@
#elif defined(__riscv)
// quants.c
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
// repack.cpp
#define ggml_quantize_mat_q8_0_4x1_generic ggml_quantize_mat_q8_0_4x1
#define ggml_quantize_mat_q8_0_4x4_generic ggml_quantize_mat_q8_0_4x4
@@ -243,6 +248,7 @@
#define quantize_row_q8_K_generic quantize_row_q8_K
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q1_0_q8_0_generic ggml_vec_dot_q1_0_q8_0
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
#define ggml_vec_dot_tq1_0_q8_K_generic ggml_vec_dot_tq1_0_q8_K
#define ggml_vec_dot_tq2_0_q8_K_generic ggml_vec_dot_tq2_0_q8_K
#define ggml_vec_dot_q2_K_q8_K_generic ggml_vec_dot_q2_K_q8_K
@@ -306,6 +312,7 @@
#define ggml_vec_dot_mxfp4_q8_0_generic ggml_vec_dot_mxfp4_q8_0
#define ggml_vec_dot_nvfp4_q8_0_generic ggml_vec_dot_nvfp4_q8_0
#define ggml_vec_dot_q1_0_q8_0_generic ggml_vec_dot_q1_0_q8_0
#define ggml_vec_dot_q2_0_q8_0_generic ggml_vec_dot_q2_0_q8_0
// repack.cpp
#define ggml_quantize_mat_q8_0_4x4_generic ggml_quantize_mat_q8_0_4x4
#define ggml_quantize_mat_q8_0_4x8_generic ggml_quantize_mat_q8_0_4x8