ggml-org/ggml

Pull Requests Commits

sync : whisper.cpp

ggerganov committed 1 year ago

Verified ea50f7ed

sync : llama.cpp

ggerganov committed 1 year ago

e3b76466

ggml : remove trailing whitespace (llama/0)

ggerganov committed 1 year ago

b44a3c3f

opencl : add GEGLU, REGLU, SWIGLU (llama/14456)

lhez committed 1 year ago

bade0d3a

Add Conv2d for CPU (llama/14388)

am17an committed 1 year ago

32e3abd2

metal : disable fast-math for some cpy kernels (llama/14460)

ggerganov committed 1 year ago

7213150b

ggml-cpu: sycl: Re-enable exp f16 (llama/14462)

Rbiessy committed 1 year ago

ff48ac98

test-backend-ops : disable llama test (llama/14461)

slaren committed 1 year ago

52b2de96

cmake : Remove redundant include path in CMakeLists.txt (llama/14452)

xiaobing318 committed 1 year ago

5b5491d5

scripts : make the shell scripts cross-platform (llama/14341)

vedranmiletic committed 1 year ago

022974b1

SYCL: disable faulty fp16 exp kernel (llama/14395)

qnixsynapse committed 1 year ago

35628677

ggml : fix unmerged GGML_FPxx_TO_FPxx refactoring (llama/14443)

CISC committed 1 year ago

6131b1ff

ggml : implement REGLU/GEGLU/SWIGLU ops (llama/14158)

CISC committed 1 year ago

2560b6c5

vulkan: Add fusion support for RMS_NORM+MUL (llama/14366)

jeffbolznv committed 1 year ago

6f5b1021

CUDA: add bf16 and f32 support to cublas_mul_mat_batched (llama/14361)

am17an committed 1 year ago

67fc917a

vulkan: handle noncontig in the final case of ggml_vk_get_cpy_pipeline (llama/14378)

jeffbolznv committed 1 year ago

6948c1f4

vulkan: lock accesses of pinned_memory vector (llama/14333)

jeffbolznv committed 1 year ago

c78c3c19

fix async_mode bug (llama/14432)

bachelor-dou committed 1 year ago

b914eb64

vulkan: Fix GGML_VULKAN_SHADER_DEBUG_INFO (llama/14427)

jeffbolznv committed 1 year ago

c5d9c4a7

ggml : add ggml_set_rows (llama/14274)

rgerganov committed 1 year ago

a1938493

cmake: regen vulkan shaders when shaders-gen sources change (llama/14398)

mtmcp committed 1 year ago

f232edff

metal : add special-case mat-vec mul for ne00 == 4 (llama/14385)

ggerganov committed 1 year ago

50be1ba6

metal : batch rows copy in a single threadgroup (llama/14384)

ggerganov committed 1 year ago

3f8162a2

musa: enable fp16 mma (all) and cublas on qy2 (llama/13842)

yeahdongcn committed 1 year ago

b6f944cb

ggml-cpu: enable IBM NNPA Vector Intrinsics (llama/14317)

taronaeo committed 1 year ago

b9061570

ggml : do not output unprintable characters on GGUF load failure (llama/14381)

CISC committed 1 year ago

08c0a1f0

sycl: GGML_SYCL_DISABLE_OPT on by default for all Intel Devices (llama/13973)

ShanoToni committed 1 year ago

4949df4f

opencl: ref count `ggml_backend_opencl_context` and refactor profiling (llama/14254)

lhez committed 1 year ago

95e4c4c2

CUDA/HIP: optimize mmv paths taken for HIP devices (llama/14324)

IMbackK committed 1 year ago

648fea5c

CUDA: mul_mat_v support for batch sizes > 1 (llama/14262)

JohannesGaessler committed 1 year ago

472b6335

Older