ochafik/llama.cpp

Pull Requests Commits
Update convert.py
ochafik committed 1 year ago
d27908c7
fast-quant: fix fopen mode in writable case (r+b)
ochafik committed 1 year ago
a64858a8
Update llama.cpp
ochafik committed 1 year ago
59d3067a
fast-quant: fix fopen mode in writable case
ochafik committed 1 year ago
024cec3d
fast-quant: fix loading of tensors (writable mmap!)
ochafik committed 1 year ago
276f2e10
fast-quant: fix crashes from --skeleton
ochafik committed 1 year ago
ce092923
Update convert.py
ochafik committed 1 year ago
8ce8c240
fast-quant: add skeleton & single tensor quant modes
ochafik committed 1 year ago
246d622f
add gguf_get_tensor_nbytes (factored out ggml_nbytes_impl & ggml_fill_nb helpers)
ochafik committed 1 year ago
71a062c5
readme : remove stray double quote (#7310)
danbev committed 1 year ago
Verified 8f7080bf
ggml : use dynamic thread scheduling for matrix multiplication (#6915)
kunnis committed 1 year ago
Verified e1b40ac3
Avoid unnecessarily disabling CUDA graphs (#7302)
agray3 committed 1 year ago
Verified dc020985
ggml : tag ggml_tensor::backend as deprecated (#7290)
slaren committed 1 year ago
Verified 344f9126
Add missing " (#7303)
AidanBeltonS committed 1 year ago
Verified 9a17ab91
embedding : free the batch after execution (#7297)
dm4 committed 1 year ago
Verified ea3b0590
sync : ggml
ggerganov committed 1 year ago
Verified 29499bb5
ggml : add `ggml_upscale_ext` (ggml/814)
balisujohn committed 1 year ago
Verified 48aa8fd1
server bench: fix bench not waiting for model load (#7284)
JohannesGaessler committed 1 year ago
Verified 583fd6b0
script : sync ggml-rpc
ggerganov committed 1 year ago
Verified 9f773486
metal : support FA without mask + add asserts (#7278)
ggerganov committed 1 year ago
Verified e8a7fd4f
sync : ggml
ggerganov committed 1 year ago
a5e3fde8
metal : tune soft_max number of threads (whisper/0)
ggerganov committed 1 year ago
f308ea70
ggml : try fix ppc64 (whisper/0)
ggerganov committed 1 year ago
c3c88f29
ggml : expose SSE3 and SSSE3 for MSVC when AVX is available (whisper/2128)
przemoc committed 1 year ago
182adefc
ggml : optimize for ppc64le using VSX intrinsics (ggml/784)
penghongbo committed 1 year ago
0d26d8cc
server: free sampling contexts on exit (#7264)
stevegrubb committed 1 year ago
Verified 4f026363
Revert "move ndk code to a new library (#6951)" (#7282)
mofosyne committed 1 year ago
Verified 1265c670
ggml : add RPC backend (#6829)
rgerganov committed 1 year ago
Verified 5e31828d
llama : disable pipeline parallelism with nkvo (#7265)
slaren committed 1 year ago
Verified 54160020
move ndk code to a new library (#6951)
eltonkola committed 1 year ago
Verified efc8f767
Terms Of Use Privacy Policy Refund Policy Impressum