vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml-cpu: support IQ4_NL_4_4 by runtime repack (#10541)

Browse source 
* ggml-cpu: support IQ4_NL_4_4 by runtime repack

* ggml-cpu: add __ARM_FEATURE_DOTPROD guard
This commit is contained in:
Shupei Fan 2024-11-28 20:52:03 +08:00 • committed by GitHub
parent 2025fa67e9
commit c202cef168
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 352 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml/include/ggml-cpu.h
View file

@ -91,6 +91,7 @@ extern "C" {
GGML_BACKEND_API int ggml_cpu_has_neon (void);
GGML_BACKEND_API int ggml_cpu_has_arm_fma (void);
GGML_BACKEND_API int ggml_cpu_has_fp16_va (void);
GGML_BACKEND_API int ggml_cpu_has_dotprod (void);
GGML_BACKEND_API int ggml_cpu_has_matmul_int8(void);
GGML_BACKEND_API int ggml_cpu_has_sve (void);
GGML_BACKEND_API int ggml_cpu_get_sve_cnt (void); // sve vector length in bytes

3
ggml/include/ggml.h
View file

@ -389,6 +389,9 @@ extern "C" {
GGML_TYPE_Q4_0_8_8 = 33,
GGML_TYPE_TQ1_0 = 34,
GGML_TYPE_TQ2_0 = 35,
GGML_TYPE_IQ4_NL_4_4 = 36,
// GGML_TYPE_IQ4_NL_4_8 = 37,
// GGML_TYPE_IQ4_NL_8_8 = 38,
GGML_TYPE_COUNT,
};