ggml-org/ggml

Pull Requests Commits

sync : llama.cpp

ggerganov committed 1 year ago

Verified c6202093

CANN: Support MOE Model MUL_MAT_ID (llama/13042)

noemotiovon committed 1 year ago

Verified 6e098c4f

cmake: use the current build config for vulkan-shaders-gen (llama/13595)

giladgd committed 1 year ago

Verified 304e2887

vulkan: move common FA code to flash_attn_base.comp (llama/13556)

jeffbolznv committed 1 year ago

Verified a2edecaf

vulkan: use scalar FA rather than coopmat2 when N==1 (llama/13554)

jeffbolznv committed 1 year ago

Verified f43e37a7

metal : add FA-vec kernel for head size 64 (llama/13583)

ggerganov committed 1 year ago

Verified 71d440de

sycl : fixed compilation warnings (llama/13582)

lslusarczyk committed 1 year ago

Verified ebf70b51

gguf : use ggml log system (llama/13571)

slaren committed 1 year ago

Verified f50f7b27

sycl: simplify bin_bcast_kernel (llama/13383)

AD2605 committed 1 year ago

Verified 548aa774

sycl: reordered Q4_K MMVQ (llama/13109)

sgeor255 committed 1 year ago

Verified 48453fb0

sycl: use oneDNN for matrices multiplication (llama/12972)

lslusarczyk committed 1 year ago

Verified 1c3eb0f2

arm64: optimize q6_k_q8_k kernel with i8mm (llama/13519)

cyb70289 committed 1 year ago

Verified 70587980

CUDA: fix crash on large batch size for quant. MoE (llama/13537)

JohannesGaessler committed 1 year ago

Verified 0c83fb7e

CUDA: faster Deepseek FA, add Turing support (llama/13435)

JohannesGaessler committed 1 year ago

Verified d587f85e

cmake: simplify vulkan shader test logic (llama/13263)

mtmcp committed 1 year ago

Verified d1f10bab

vulkan: KHR_coopmat flash attention (llama/13506)

jeffbolznv committed 1 year ago

Verified 9cccb7af

vulkan: workaround FA compile failures on macos (llama/13517)

jeffbolznv committed 1 year ago

Verified 70c74ea5

metal : use FA-vec kernel up to batch size 20 (llama/13496)

ggerganov committed 1 year ago

Verified 01732156

metal : optimize multi-sequence FA vec kernel (llama/13493)

ggerganov committed 1 year ago

Verified 5c8105e4

ggml-cpu: Update KleidiAI to v1.6 and fix include directives (llama/13509)

eddnjjn committed 1 year ago

Verified 6fe93b94

mnist: fix segmentation fault (#1227)

JohannesGaessler committed 1 year ago

Verified 7c06c10c

ggml : fix apple OS check in ggml_print_backtrace (#1229)

slaren committed 1 year ago

Verified cc309623

ggml : Fix missing backtrace on Linux (#1228)

danielzgtg committed 1 year ago

Verified 41ebdd82

sync : whisper.cpp

ggerganov committed 1 year ago

9b048bb7

examples : update link to Paul Tol's color scheme [no ci] (whisper/3140)

danbev committed 1 year ago

4276fbd5

examples : update to ggml-opt and ggml-backend changes (#0)

ggerganov committed 1 year ago

148b2863

sync : llama.cpp

ggerganov committed 1 year ago

e219fb3c

opencl: remove unnecessary assert for `add` (llama/13257)

lhez committed 1 year ago

76176a22

llama/ggml: add LLM training support (llama/10544)

JohannesGaessler committed 1 year ago

a0e6394b

ggml-cpu: Integrate fp32=bf16xbf16 SME KleidiAI kernel (llama/13053)

eddnjjn committed 1 year ago

51e1afc6

Older