ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-24 08:29:29 +00:00
Code Issues Packages Projects Releases Wiki Activity

mxfp4: basics

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-08-08 08:41:04 +03:00
parent 58f3bda0ae
commit 58c3bffff4
7 changed files with 215 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
ggml/include/ggml.h
View File

@@ -403,6 +403,7 @@ extern "C" {
GGML_TYPE_Q4_0_4_4 = 31,
GGML_TYPE_Q4_0_4_8 = 32,
GGML_TYPE_Q4_0_8_8 = 33,
GGML_TYPE_MXFP4 = 39, // so we are compatible with mainline
//
// So we are able to consume MS BitNet I2_S quants
//
@@ -507,9 +508,10 @@ extern "C" {
GGML_FTYPE_MOSTLY_IQ4_XS = 22, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ1_M = 23, // except 1d tensors
GGML_FTYPE_MOSTLY_BF16 = 24, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_4_4 = 25, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_4_8 = 26, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_8_8 = 27, // except 1d tensors
GGML_FTYPE_MOSTLY_MXFP4 = 25, // except 1d tensors, using 26 to be compatible with mainline
GGML_FTYPE_MOSTLY_Q4_0_4_4 = 26, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_4_8 = 27, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_8_8 = 28, // except 1d tensors
//
GGML_FTYPE_MOSTLY_Q6_0 = 127, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ1_BN = 128, // except 1d tensors