ggml-org/llama.cpp

Pull Requests Commits
Fixed CUDA runtime version check (#1879)
JohannesGaessler committed 2 years ago
Verified a09f9195
cmake : remove whitespaces
ggerganov committed 2 years ago
Verified bed92756
examples : add chat-vicuna.sh (#1854)
yangli2 committed 2 years ago
Verified c36e81da
cmake : set include path for OpenBlas (#1830)
okigan committed 2 years ago
Verified 3559433f
swift : Package compile breaks due to ggml-metal.metal (#1831)
Schaltfehler committed 2 years ago
Verified 69b34a0e
make : add train-text-from-scratch (#1850)
daboe01 committed 2 years ago
Verified cf267d1c
readme : server compile flag (#1874)
nivibilla committed 2 years ago
Verified 9dda13e5
make : clean *.so files (#1857)
sandyiscool committed 2 years ago
Verified 37e257c4
Fix the validation of main device (#1872)
howard0su committed 2 years ago
Verified 64cc19b4
metal : parallel command buffer encoding (#1860)
ggerganov committed 2 years ago
Verified 4bfcc855
Better error when using both LoRA + GPU layers (#1861)
JohannesGaessler committed 2 years ago
Verified 6b8312e7
CUDA full GPU acceleration, KV cache in VRAM (#1827)
JohannesGaessler committed 2 years ago
Verified 254a7a7a
baby-llama : fix operator!= (#1821)
0xspringtime committed 2 years ago
Verified 92549202
train : improved training-from-scratch example (#1652)
xaedes committed 2 years ago
Verified e32089b2
llama : do a warm-up eval at start for better timings (#1824)
ggerganov committed 2 years ago
Verified 2347e45e
Allow "quantizing" to f16 and f32 (#1787)
KerfuffleV2 committed 2 years ago
Verified 74d4cfa3
Metal implementation for all k_quants (#1807)
ikawrakow committed 2 years ago
Verified 74a6d922
ci : run when changing only the CUDA sources (#1800)
slaren committed 2 years ago
Verified e4caa8da
Leverage mmap for offloading tensors to GPU (#1597)
howard0su committed 2 years ago
Verified 58970a4c
metal : fix failure to load model (#1817)
ikawrakow committed 2 years ago
Verified 8c0a10e6
Fix issue where interactive mode crashes when input exceeds ctx size (#1789)
KerfuffleV2 committed 2 years ago
Verified fa84c4b3
Fixed WSL cuda's OOM error (#1594)
JoelSeniorLiang committed 2 years ago
Verified 12b063f0
Update SHA256SUMS with current hashes for models quantized using q4_0 (#1798)
rlanday committed 2 years ago
Verified 31d2b5f4
cmake : fix Metal build (close #1791)
ggerganov committed 2 years ago
Verified 4de0334f
k-quants : GCC12 compilation fix (#1792)
vagran committed 2 years ago
Verified 3f122315
metal : fix issue with ggml-metal.metal path. Closes #1769 (#1782)
abetlen committed 2 years ago
Verified 303f5809
doc : fix wrong address of BLIS.md (#1772)
Aisuko committed 2 years ago
Verified 059e9906
ggml : force no_alloc == false when creating opt tensors (close #1699)
ggerganov committed 2 years ago
Verified 17c10acf
metal : add Q4_1 implementation (#1785)
ikawrakow committed 2 years ago
Verified e9b66ee9
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691)
KerfuffleV2 committed 2 years ago
Verified 4f0154b0
Terms Of Use Privacy Policy Refund Policy Impressum