Commits

History for

ondev-pad-reflect

Commits on Jun 26, 2024

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

Commits on Jun 15, 2024

metal : utilize max shared memory for mul_mat_id (llama/7935)
ggerganov
committedJun 15, 2024
move BLAS to a separate backend (llama/6210)

slaren
and
ggerganov
committedJun 15, 2024
tests : add non-cont unary tests (llama/7857)
ggerganov
committedJun 15, 2024
ggml : refactor rope norm/neox (llama/7634)
ggerganov
committedJun 15, 2024
CUDA: quantized KV support for FA vec (llama/7527)

JohannesGaessler
authored and
ggerganov
committedJun 15, 2024
metal : remove invalid asserts (llama/7617)
ggerganov
committedJun 15, 2024
metal : add missing asserts (llama/7617)
ggerganov
committedJun 15, 2024
ggml : fix YARN + add tests + add asserts (llama/7617)
ggerganov
committedJun 15, 2024

Commits on May 29, 2024

ggml : generalize GGML_OP_CONCAT (llama/7563)
ggerganov
committedMay 29, 2024
metal : add GGML_OP_REPEAT kernels (llama/7557)
ggerganov
committedMay 29, 2024
metal : disable FA kernel for HS=256 (llama/7556)
ggerganov
committedMay 29, 2024

Commits on May 28, 2024

ggml : drop support for QK_K=64 (llama/7473)
ggerganov
committedMay 28, 2024
llama : add phi3 128K model support (llama/7225)

liuwei-git
and
ggerganov
committedMay 28, 2024

Commits on May 15, 2024

ggml : add ggml_upscale_ext (ggerganov#814 )

balisujohn
and
ggerganov
committedMay 15, 2024

Commits on May 14, 2024

metal : support FA without mask + add asserts (llama/7278)
ggerganov
committedMay 14, 2024
metal : tune soft_max number of threads (whisper/0)
ggerganov
committedMay 14, 2024

Commits on May 11, 2024

metal : fix indent (#0)
ggerganov
committedMay 11, 2024
ggml : full ALiBi support (llama/7192)
ggerganov
committedMay 11, 2024
metal : fix flash attention kernel requirements (llama/7169)
ggerganov
committedMay 11, 2024
metal : use vm_allocate instead of posix_memalign on macOS (llama/7078)

giladgd
authored and
ggerganov
committedMay 11, 2024
ggml : introduce bfloat16 support (llama/6412)

jart
authored and
ggerganov
committedMay 11, 2024
switch to using localizedDescription (llama/7010)

bakkot
authored and
ggerganov
committedMay 11, 2024
metal : remove deprecated error code (llama/7008)
ggerganov
committedMay 11, 2024
metal : log more info on error (llama/6987)

bakkot
authored and
ggerganov
committedMay 11, 2024
ggml : add Flash Attention (llama/5021)

committedMay 11, 2024
ggml : group all experts in a single ggml_mul_mat_id (llama/6505)

slaren
and
ggerganov
committedMay 11, 2024
llama : add qwen2moe (llama/6074)

simonJJJ
and
ggerganov
committedMay 11, 2024
Added support for GGML_OP_CLAMP in Metal (llama/6662)

authored and
ggerganov
committedMay 11, 2024
metal : unify mul_mv_id kernels (llama/6556)

slaren
authored and
ggerganov
committedMay 11, 2024

Commits on May 1, 2024

feat: implemented sigmoid function (ggerganov#806 )
justcho5
committedMay 1, 2024

Commits on Apr 6, 2024

ggml : mul_mat_id use the same tensor for all the experts (llama/6387)

slaren
and
ggerganov
committedApr 6, 2024

Commits on Mar 27, 2024

llama : greatly reduce output buffer memory usage (llama/6122)

compilade
and
ggerganov
committedMar 27, 2024
IQ1_M: 1.75 bpw quantization (llama/6302)

authored and
ggerganov
committedMar 27, 2024
metal : proper assert for mat-mat memory alignment (llama/6225)
ggerganov
committedMar 27, 2024