ggerganov/llama.cpp

Pull Requests Commits

Vectorize q load

Aidan committed 1 year ago

a235b7c5

Store scales in local mem

Aidan committed 1 year ago

604ef6bf

Single load for half2

Aidan committed 1 year ago

cb3fb420

Remove double lines

Aidan committed 1 year ago

4a481556

Merge pull request #7920 from ggerganov/codeplay/revert-host-alloc

joeatodd committed 1 year ago

Verified ff076b88

Merge pull request #7919 from ggerganov/codeplay/unify-rope-sycl

joeatodd committed 1 year ago

Verified b2c8c831

Replace powf with sycl::pow in ggml-sycl.cpp

joeatodd committed 1 year ago

ded54b5d

Revert "use the correct SYCL context for host USM allocations"

joeatodd committed 1 year ago

18133cab

joeatodd committed 1 year ago

abd7c7b8

[SYCL] Update unsupported ops

joeatodd committed 1 year ago

0c0f3f00

[SYCL] unify rope norm/neox

joeatodd committed 1 year ago

9b81b572

tests : add non-cont unary tests (#7857)

ggerganov committed 1 year ago

Verified a9cae480

ggml : improve ggml_is_contiguous logic (#7856)

ggerganov committed 1 year ago

Verified bfaa676b

server : restore numeric prompts (#7883)

ggerganov committed 1 year ago

Verified 704a35b1

update intel docker oneapi-basekit to 2024.1.1-devel-ubuntu22.04 (#7894)

airMeng committed 1 year ago

Verified dcf75270

Fix a typo and add Fedora 40 pacakge to install for Vulkan (#7794) [no ci]

metal3d committed 1 year ago

Verified f2b5764b

vulkan: select only one device for single gpu with multiple drivers (#7582)

Adriankhl committed 1 year ago

Verified 73bac2b1

Update Vulkan RoPE implementation (#7818)

0cc4m committed 1 year ago

Verified ef52d1d1

fix broken link in pr template (#7880) [no ci]

deven367 committed 1 year ago

Verified 14f83526

github: move PR template to .github/ root (#7868)

mofosyne committed 1 year ago

Verified 6fe42d07

llama-bench: more compact markdown tables (#7879)

JohannesGaessler committed 1 year ago

Verified 148995e5

tests : check the Python version (#7872)

ggerganov committed 1 year ago

Verified 4bfe50f7

CUDA: int8 tensor cores for MMQ (q4_K, q5_K, q6_K) (#7860)

JohannesGaessler committed 1 year ago

Verified bdcb8f42

fix CUDA CI by using a windows-2019 image (#7861)

slaren committed 1 year ago

Verified c2ce6c47

json: refine constraint for whitespace to avoid runaways yet allow pretty print (#7866)

ochafik committed 1 year ago

Verified b61eb964

`json`: document schema conversion in GBNF readme, align manual grammar examples & converters (#7841)

ochafik committed 1 year ago

Verified 396b18df

cmake : fix CMake requirement for CUDA (#7821)

cebtenzzre committed 1 year ago

Verified 864a99e7

ci : try win-2019 on server windows test (#7854)

slaren committed 1 year ago

Verified fd5ea0f8

examples : remove --instruct remnants (#7846)

ggerganov committed 1 year ago

Verified c28a8390

server : improve "prompt" handling (#7847)

ggerganov committed 1 year ago

Verified d9da0e49

Older