cosmopolitan/third_party/ggml
Justine Tunney fa20edc44d
Reduce header complexity
- Remove most __ASSEMBLER__ __LINKER__ ifdefs
- Rename libc/intrin/bits.h to libc/serialize.h
- Block pthread cancelation in fchmodat() polyfill
- Remove `clang-format off` statements in third_party
2023-11-28 14:39:42 -08:00
..
BUILD.mk Rename makefiles BUILD.mk 2023-11-28 11:21:08 -08:00
common.cc Reduce header complexity 2023-11-28 14:39:42 -08:00
common.h Reduce header complexity 2023-11-28 14:39:42 -08:00
companionai.txt Upgrade llama.cpp to e6a46b0ed1884c77267dc70693183e3b7164e0e0 2023-05-10 04:20:48 -07:00
fp16.c Reduce header complexity 2023-11-28 14:39:42 -08:00
fp16.h Reduce header complexity 2023-11-28 14:39:42 -08:00
fp16.internal.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.internal.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q4_0.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q4_0.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q4_1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q4_1.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q4_2.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q4_2.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q5_0.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q5_0.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q5_1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q5_1.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q8_0.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q8_0.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q8_1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v1.q8_1.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.internal.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q4_0.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q4_0.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q4_1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q4_1.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q5_0.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q5_0.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q5_1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q5_1.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q8_0.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q8_0.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q8_1.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggjt.v2.q8_1.h Reduce header complexity 2023-11-28 14:39:42 -08:00
ggml.c Reduce header complexity 2023-11-28 14:39:42 -08:00
ggml.h Reduce header complexity 2023-11-28 14:39:42 -08:00
LICENSE Import llama.cpp 2023-04-27 14:37:14 -07:00
llama.cc Reduce header complexity 2023-11-28 14:39:42 -08:00
llama.h Reduce header complexity 2023-11-28 14:39:42 -08:00
llama_util.h Reduce header complexity 2023-11-28 14:39:42 -08:00
main.cc Reduce header complexity 2023-11-28 14:39:42 -08:00
perplexity.cc Reduce header complexity 2023-11-28 14:39:42 -08:00
quantize.cc Reduce header complexity 2023-11-28 14:39:42 -08:00
README.cosmo Introduce support for GGJT v3 file format 2023-06-03 15:46:21 -07:00

DESCRIPTION

  ggml is a machine learning library useful for LLM inference on CPUs

LICENSE

  MIT

ORIGIN

  https://github.com/ggerganov/llama.cpp
  d8bd0013e8768aaa3dc9cfc1ff01499419d5348e

LOCAL CHANGES

  - Maintaining support for deprecated file formats
  - Make it possible for loaded prompts to be cached to disk
  - Introduce -v and --verbose flags
  - Reduce batch size from 512 to 32
  - Allow --n_keep to specify a substring of prompt
  - Don't print stats / diagnostics unless -v is passed
  - Reduce --top_p default from 0.95 to 0.70
  - Change --reverse-prompt to no longer imply --interactive
  - Permit --reverse-prompt specifying custom EOS if non-interactive
  - Refactor headers per cosmo convention
  - Remove C++ exceptions; use Die() function instead
  - Removed division from matrix multiplication.
  - Let quantizer convert between ggmt formats