Improve Vulkan shader build system (llama/9239)
9746f77
Markus Tavenrath
commited on
Spaces:
Sleeping
Sleeping
Commit History
ggml-quants : ternary packing for TriLMs and BitNet b1.58 (llama/8151)
d1c244a
cuda : fix defrag with quantized KV (llama/9319)
061ca37
slaren
commited on
ggml : AVX2 support for Q4_0_8_8 (llama/8713)
480ad4d
Srihari-mcw
commited on
Fix DMMV dequantization (llama/9279)
aa12d61
Ouadie EL FAROUKI
commited on
ggml : add pthread includes on FreeBSD (llama/9258)
d558e0e
yuri@FreeBSD
commited on
Threadpool: take 2 (llama/8672)
e3e9ca4
vulkan: fix compilation with GGML_VULKAN_DEBUG=ON (ggml/948)
7f60aae
vulkan: add dryrun support to sin and cos ops (ggml/947)
e2fe267
vulkan: correctly report support for OP_CONT (ggml/946)
098f7fa
tests: add gradient tests for all backends (ggml/932)
4751b2f
go : add temperature options (#2417)
5b36f0b
unverified
Binozo
Binozo
commited on
docker : add libsdl2-dev for container builds (#2424)
aa93432
unverified
JohnnyB
commited on
go : add tests and update bindings (#2425)
c80d17a
unverified
Stavros Panakakis
commited on
server : use OS-generated temp file name for converted files (#2419)
04d9c8d
unverified
go : fix CUDA build (#2416)
dafe96d
unverified
Binozo
Binozo
commited on
cann : add Ascend NPU instructions (#2410)
ae9acd3
unverified
cmake: Fix libdir value in pkgconfig file (#2407)
a048ef3
unverified
Philippe Normand
commited on
revert : cmake : set MSVC to use UTF-8 on source files (#2346)
5e9ff52
sync : ggml
b13db51
ggml: fix ggml_graph_cpy undefined behavior (ggml/943)
9202e70
cann : fix doxy (ggml/0)
406ac07
vulkan : fix build (llama/0)
e237370
cuda : mark BF16 CONT as unsupported
561bebd
cmake : set MSVC to use UTF-8 on source files (#2346)
9b3df8e
unverified
Tim Miller
commited on
readme : remove invalid flag from Python example (#2396)
5372e8b
unverified
UsernamesLame
commited on
readme : fix link (#2394)
ae51c50
unverified
go : add beamsize/entropythold/maxcontext to context interface (#2350)
7efcda7
unverified
hsinhoyeh
commited on
talk-llama : sync llama.cpp
4493ffd
whisper : update FA call
2bfec97
sync : ggml
7ba8c97
sync : vulkan (skip) (llama/0)
5fe3dd6
ggml : do not crash when quantizing q4_x_x with an imatrix (llama/9192)
d64f932
slaren
commited on
metal : separate scale and mask from QKT in FA kernel (llama/9189)
90cc3cd
ggml : add SSM Metal kernels (llama/8546)
b6e7294
metal : gemma2 flash attention support (llama/9159)
e62fd15
slaren
commited on
CPU/CUDA: Gemma 2 FlashAttention support (llama/8542)
fb8ae8b
Add a space to supress a cmake warning (llama/9133)
287612e
Add oneDNN primitive support (llama/9091)
b4d8c3e
fallback mmvq (llama/9088)
4b1fda0
Fix SYCL `im2col` and `convert` Overflow with Large Dims (llama/9052)
5f43886
rpc : print error message when failed to connect endpoint (llama/9042)
d54b156
rpc : prevent crashes on invalid input (llama/9040)
656ae00
ggml : dynamic ggml_sched_max_splits based on graph_size (llama/9047)
e0dc1ad
cmake : remove unused option GGML_CURL (llama/9011)
12634fc
ggml : move rope type enum to ggml.h (llama/8949)
9d45f48
ggml: fix div-by-zero (llama/9003)
d9ee26f
DavidKorczynski
commited on