ggerganov/llama.cpp

Pull Requests Commits

Metal: faster Q4_0 and Q4_1 matrix x vector kernels (#2212)

ikawrakow committed 2 years ago

Verified 27ad57a6

Revert "Support using mmap when applying LoRA (#2095)" (#2206)

howard0su committed 2 years ago

Verified 32c54116

Fix compile error on Windows CUDA (#2207)

howard0su committed 2 years ago

Verified ff5d58fa

devops : add missing quotes to bash script (#2193)

bodograumann committed 2 years ago

Verified b782422a

metal : new q4_0 matrix-vector kernel (#2188)

lshzh-ww committed 2 years ago

Verified 1cbf5614

ggml : broadcast mul_mat + conv batch support (#2199)

ggerganov committed 2 years ago

Verified 975221e9

ggml : add ggml_pool_1d and ggml_pool_2d

ggerganov committed 2 years ago

4523d10d

cuda : add gelu support

ggerganov committed 2 years ago

680e6f91

FP16 is supported in CM=6.0 (#2177)

howard0su committed 2 years ago

Verified 4e7464ef

Fixed __dp4a compute capability: 6.0 -> 6.1 (#2189)

JohannesGaessler committed 2 years ago

Verified 2b5eb72e

ggml : revert CUDA broadcast changes from #2183 (#2191)

ggerganov committed 2 years ago

Verified f7d278fa

ggml : sync (abort callback, mul / add broadcast, fix alibi) (#2183)

ggerganov committed 2 years ago

Verified 20d7740a

ggml : remove src0 and src1 from ggml_tensor and rename opt to src (#2178)

spencersutton committed 2 years ago

Verified 5bf2a277

llama : add classifier-free guidance (#2135)

bullno1 committed 2 years ago

Verified c9c74b4e

docker : add '--server' option (#2174)

williamjeong2 committed 2 years ago

Verified 3ec7e596

readme : fix zig build instructions (#2171)

chadbrewbaker committed 2 years ago

Verified 917831c6

Support using mmap when applying LoRA (#2095)

howard0su committed 2 years ago

Verified 23474632

Possible solution to allow K-quants on models with n_vocab!=32000 (#2148)

LostRuins committed 2 years ago

Verified bbef2821

mpi : add support for distributed inference via MPI (#2099)

evanmiller committed 2 years ago

Verified 5656d105

llama : remove "first token must be BOS" restriction (#2153)

oobabooga committed 2 years ago

Verified 1d163099

main : escape prompt prefix/suffix (#2151)

pnb committed 2 years ago

Verified db4047ad

readme : update Termux instructions (#2147)

JackJollimore committed 2 years ago

Verified 18780e0a

ggml : fix buidling with Intel MKL but ask for "cblas.h" issue (#2104) (#2115)

clyang committed 2 years ago

Verified 3bbc1a11

readme : add more docs indexes (#2127)

rankaiyx committed 2 years ago

Verified 2492a53f

Fixed OpenLLaMA 3b CUDA mul_mat_vec_q (#2144)

JohannesGaessler committed 2 years ago

Verified 64639555

CUDA: add __restrict__ to mul mat vec kernels (#2140)

JohannesGaessler committed 2 years ago

Verified 061f5f8d

docker : add support for CUDA in docker (#1461)

canardleteer committed 2 years ago

Verified 84525e79

ci : switch threads to 1 (#2138)

ggerganov committed 2 years ago

Verified a7e20edf

ggml : change ggml_graph_compute() API to not require context (#1999)

mqy committed 2 years ago

Verified 1d656d63

ggml : remove sched_yield() call in ggml_graph_compute_thread() (#2134)

ggerganov committed 2 years ago

72421402

Newer Older