ggml-org/llama.cpp

Pull Requests Commits

allocator: cleanup, more comments

slaren committed 2 years ago

d273bfd2

llama.cpp: print input/output buffers size

slaren committed 2 years ago

5141472e

allocator cleanup

slaren committed 2 years ago

e2b95759

allocator: fix partial offloading

slaren committed 2 years ago

7de78825

allocator: automatic inplace operations

slaren committed 2 years ago

e87840f9

improved memory management fixes

slaren committed 2 years ago

3d679827

llama.cpp: partially restore state support, graph export

slaren committed 2 years ago

56e9ae06

remove unused code

slaren committed 2 years ago

37d3f6a2

improved memory management

slaren committed 2 years ago

cd6f5dec

initial implementation of delayed graph allocation

slaren committed 2 years ago

de69f8f2

automatically calculate compute buffer sizes (without graph allocator)

slaren committed 2 years ago

cb205c0d

llama.cpp: remove backend-specific code where possible

slaren committed 2 years ago

77ac8dea

slaren committed 2 years ago

295f8565

fix double-free with --no-mmap

slaren committed 2 years ago

1102ff56

improve layer backend printing with ranges

slaren committed 2 years ago

4e94af30

only allocate as much memory as is required in each backend for the model

slaren committed 2 years ago

c2beeb8e

rebase to master (except ggml-cuda)

slaren committed 2 years ago

9c72e7e9

fix NVCC version on Makefile, __halves2half2 -> make_half2

slaren committed 2 years ago

33ab185d

slaren committed 2 years ago

24cc6f00

restore simple.cpp for now

slaren committed 2 years ago

5765d7a5

ggml backend interface wip

slaren committed 2 years ago

0d2b66c6

llama : add custom RoPE (#2054)

jxy committed 2 years ago

Verified 6e7cca40

flake : add runHook preInstall/postInstall to installPhase so hooks function (#2224)

ddellacosta committed 2 years ago

Verified a6803cab

make : use pkg-config for OpenBLAS (#2222)

Freed-Wu committed 2 years ago

Verified 7dabc66f

cuda : allocate all temporary ggml_tensor_extra_gpu from a fixed-size buffer (#2220)

bullno1 committed 2 years ago

Verified 7cdd30bf

ggml : fix static_assert with older compilers #2024 (#2218)

evanmiller committed 2 years ago

Verified e8035f14

llama : add functions that work directly on model (#2197)

bullno1 committed 2 years ago

Verified 7513b7b0

build.zig : install config header (#2216)

alichraghi committed 2 years ago

Verified de834242

examples : fixed path typos in embd-input (#2214)

xushangning committed 2 years ago

Verified c48c525f

cuda : support broadcast add & mul (#2192)

li-plus committed 2 years ago

Verified 206e01de

Older