Update llama.cpp

This commit is contained in:
Andrei Betlen 2024-01-30 09:48:09 -05:00
parent da003d8768
commit 011cd84ded
3 changed files with 9 additions and 1 deletions

View file

@ -30,6 +30,12 @@ build.metal:
build.vulkan:
CMAKE_ARGS="-DLLAMA_VULKAN=on" python3 -m pip install --verbose -e .
build.kompute:
CMAKE_ARGS="-DLLAMA_KOMPUTE=on" python3 -m pip install --verbose -e .
build.sycl:
CMAKE_ARGS="-DLLAMA_SYCL=on" python3 -m pip install --verbose -e .
build.sdist:
python3 -m build --sdist

View file

@ -187,6 +187,7 @@ LLAMA_TOKEN_TYPE_BYTE = 6
# LLAMA_FTYPE_MOSTLY_IQ2_XS = 20, // except 1d tensors
# LLAMA_FTYPE_MOSTLY_Q2_K_S = 21, // except 1d tensors
# LLAMA_FTYPE_MOSTLY_Q3_K_XS = 22, // except 1d tensors
# LLAMA_FTYPE_MOSTLY_IQ3_XXS = 23, // except 1d tensors
# LLAMA_FTYPE_GUESSED = 1024, // not specified in the model file
# };
@ -211,6 +212,7 @@ LLAMA_FTYPE_MOSTLY_IQ2_XXS = 19
LLAMA_FTYPE_MOSTLY_IQ2_XS = 20
LLAMA_FTYPE_MOSTLY_Q2_K_S = 21
LLAMA_FTYPE_MOSTLY_Q3_K_XS = 22
LLAMA_FTYPE_MOSTLY_IQ3_XXS = 23
LLAMA_FTYPE_GUESSED = 1024
# enum llama_rope_scaling_type {

2
vendor/llama.cpp vendored

@ -1 +1 @@
Subproject commit 2aed77eb06a329f0d82bb1c467f4244904d4073f
Subproject commit 8f8ddfcfadc830b936318c3ea9fe2e8e3365aa85