ggml : fix fallback to CPU for ununsupported ops (llama/15118)
2b7ae5e
Diego Devesa
commited on
Spaces:
Running
Running
Commit History
CANN: add support for ACL Graph (llama/15065)
137a0dc
Chenguang Li
commited on
sycl: fix mul_mat selection (llama/15092)
344310a
Romain Biessy
commited on
cmake: Add GGML_BACKEND_DIR option (llama/15074)
6e460b6
Christian Kastner
commited on
vulkan: fix build when using glslang that does not support coopmat2 (llama/15062)
863e083
vulkan: Use coopmat2 for conv2d (llama/14982)
6df82f4
opencl: fix adreno compiler detection logic (llama/15029)
e6a209e
lhez
commited on
CUDA: use mma FA kernel for gqa > 4 on RTX 4000 (llama/15035)
9e85264
cuda: make im2col a little faster (llama/15025)
9a85c65
cuda, sycl : fix batched gemm when ne02 == 1 && ne03 > 1 (llama/15038)
cc3a2ed
vulkan: coopmat2 mul_mat optimizations (llama/14934)
ca86566
vulkan: Support ne[3]>1 in noncontig matrix-vector multiply (llama/15015)
d4c4115
vulkan: optimizations for direct convolution (llama/14933)
215f463
CUDA: fix MMQ nwarps for AMD with warp_size==32 (llama/15014)
fbc3cd1
opencl: add f16 for `add`, `sub`, `mul`, `div` (llama/14984)
4dc1834
lhez
commited on
ggml : Q2k interleaving implementation - x86/x64 SIMD (llama/14373)
e2965b0
Vulkan: Fix minor debug mode issues (llama/14899)
a81bc86
CANN: Improve loading efficiency after converting weights to NZ format. (llama/14985)
7612978
opencl: add `mul_mat_f32_f32_l4_lm` and `mul_mat_f16_f32_l4_lm` (llama/14809)
05577c3
lhez
commited on
HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes (llama/14949)
149f5a5
CUDA: skip masked KV slices for all FA kernels (llama/14924)
0c60f80
HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (llama/14945)
e37eff3
HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (llama/14930)
f9dbd96
HIP: Ignore unsupported unroll transformation in fattn-vec (llama/14931)
8e133f7
CANN: Add ggml_set_rows (llama/14943)
fa22f70
cuda : add softcap fusion (llama/14907)
2237878
Sigbjørn Skjæret
commited on
CUDA: add roll (llama/14919)
d41a4ec
ggml-cpu : deduplicate scalar implementations (llama/14897)
1d58d7c
xctan
commited on
SYCL: Add set_rows support for quantized types (llama/14883)
c55b72b
Akarshan Biswas
commited on
CUDA: fix pointer incrementation in FA (llama/14916)
eb84e7e
sycl: refactor quantization to q8_1 (llama/14815)
31edd77
Alberto Cabrera Pérez
commited on
cmake : Fix BLAS link interface (ggml/1316)
3020711
Kai Pastor
commited on
vulkan : fix 32-bit builds (ggml/1313)
96b66fd
Kai Pastor
commited on
scripts : update sync scripts
311eccd
node : add win platform check for require path (#3363)
29b8653
unverified
ci : update main-cuda.Dockerfile (#3371)
e79709c
unverified
ustas
commited on
whisper : fixed crash in GPU device selection on multi-GPU systems (#3372)
0869200
unverified
Dw9
commited on
wasm : change ggml model host to HF (#3369)
ac86ad0
unverified
ruby : Add ruby binding for max_len (#3365)
b408b8e
unverified
Adam Debono
commited on
stream.wasm : add language selection support (#3354)
e8933c1
unverified
whisper : reset conv scheduler when CoreML is used (#3350)
f425556
unverified
ggml : remove old kompute, cann (skip) (#3349)
d321914
unverified
talk-llama : sync llama.cpp
844e617
sync : ggml
7d38d31
vulkan : add fp16 support for the conv_2d kernel (llama/14872)
48e92ad
Erik Scholz
commited on
vulkan: skip empty set_rows to avoid invalid API usage (llama/14860)
22fb24a
HIP: Enable Matrix cores for MMQ Kernels, Enable stream-K for CDNA 3 (llama/14624)
5422b31
deepsek
commited on