Commits

History for

ondev-pad-reflect

Commits on Jun 29, 2024

Commits on Jun 26, 2024

ggml : add GGML_CUDA_USE_GRAPHS option, restore GGML_CUDA_FORCE_CUBLAS (cmake) (llama/8140)

slaren
authored and
ggerganov
committedJun 26, 2024
ggml : reorganize source code + improve CMake (ggerganov#865 )
ggerganov
committedJun 26, 2024

Commits on Jun 18, 2024

whisper : use ggml_backend_sched (whisper/2239)

ggerganov
and
slaren
committedJun 18, 2024

Commits on Jun 16, 2024

cuda : fix bounds check for src0 rows in MMVQ kernel (whisper/2231)

ggerganov
and
JohannesGaessler
committedJun 16, 2024
move BLAS to a separate backend (cont) (llama/6210)

slaren
authored and
ggerganov
committedJun 16, 2024
Vulkan Shader Refactor, Memory Debugging Option (llama/7947)

0cc4m
authored and
ggerganov
committedJun 16, 2024
ggml : remove OpenCL (#0)
ggerganov
committedJun 16, 2024
cmake : fix cuda vars (#0)
ggerganov
committedJun 16, 2024
ggml : fix and optimize ppc64le (ggerganov#849 )
penghongbo
committedJun 16, 2024
ggml : remove duplicate include of ggml-common.h (ggerganov#853 )
danbev
committedJun 16, 2024
sycl : remove global variables (cont) (llama/7710)

Nuullll
and
airMeng
committedJun 16, 2024

Commits on Jun 15, 2024

ggml : remove opencl (#0)
ggerganov
committedJun 15, 2024
cuda : update build (#0)
ggerganov
committedJun 15, 2024
remove global variables (llama/7710)

airMeng
authored and
ggerganov
committedJun 15, 2024
CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (llama/7921)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
metal : utilize max shared memory for mul_mat_id (llama/7935)
ggerganov
committedJun 15, 2024
rpc : fix ggml_backend_rpc_supports_buft() (llama/7918)

rgerganov
authored and
ggerganov
committedJun 15, 2024
move BLAS to a separate backend (llama/6210)

slaren
and
ggerganov
committedJun 15, 2024
CUDA: fix broken oob check for FA vec f32 kernel (llama/7904)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
tests : add non-cont unary tests (llama/7857)
ggerganov
committedJun 15, 2024
ggml : improve ggml_is_contiguous logic (llama/7856)
ggerganov
committedJun 15, 2024
vulkan: select only one device for single gpu with multiple drivers (llama/7582)

Adriankhl
authored and
ggerganov
committedJun 15, 2024
Update Vulkan RoPE implementation (llama/7818)

authored and
ggerganov
committedJun 15, 2024
CUDA: int8 tensor cores for MMQ (q4_K, q5_K, q6_K) (llama/7860)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
CUDA: use tensor cores for MMQ (llama/7676)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
use the correct SYCL context for host USM allocations (llama/7777)

bashbaug
authored and
ggerganov
committedJun 15, 2024
CUDA: revise q8_1 data layout for mul_mat_q (llama/7824)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
vulkan : reuse parent extra for views (llama/7806)

authored and
ggerganov
committedJun 15, 2024
fix softmax r2r result wrong issue (llama/7811)

pengxin99
authored and
ggerganov
committedJun 15, 2024
CUDA: refactor mmq, dmmv, mmvq (llama/7716)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
ggml : refactor rope norm/neox (llama/7634)
ggerganov
committedJun 15, 2024