ggerganov/llama.cpp

Pull Requests Commits

llama : disable FA if KV head size do not match

ggerganov committed 1 year ago

Verified ef79941a

Add Nix and Flox install instructions (#7899)

bryanhonof committed 1 year ago

Verified b473e950

sched : offload_op also requires supports_op (#7977)

slaren committed 1 year ago

Verified 99052cd2

fix: divide 0 exception in mamba (#7932)

thxCode committed 1 year ago

Verified c637fcd3

Implement non-mapped async IO for CUDA on Windows. (#7896)

mtavenrath committed 1 year ago

Verified 6a2f0b34

rpc : fix load/store misaligned addresses (#7948)

ggerganov committed 1 year ago

Verified 21be9cab

gguf-dump.py: add --markdown dump output (#7853)

mofosyne committed 1 year ago

Verified 006167aa

[SYCL] Update README-sycl.md for Chapter "Recommended release" and "News" (#7946)

arthw committed 1 year ago

Verified df68d4fa

Add support for sqrt on CUDA (#7953)

calvin-laurenson committed 1 year ago

Verified 43b35e38

cuda : fix bounds check for src0 rows in MMVQ kernel (whisper/2231)

ggerganov committed 1 year ago

Verified 19b7a836

ggml : fix and optimize ppc64le (ggml/849)

penghongbo committed 1 year ago

Verified b5fcf8ef

ggml : remove duplicate include of ggml-common.h (ggml/853)

danbev committed 1 year ago

Verified 398105ff

flake.lock: Update (#7951)

ggerganov committed 1 year ago

Verified bc6c457f

unicode : avoid char32_t (#7957)

ggerganov committed 1 year ago

Verified 52399254

readme : update UI list [no ci] (#7958)

hopkins385 committed 1 year ago

Verified 6fe1c627

ggml : fix handling of zero blocks in IQ quants (#7955)

ggerganov committed 1 year ago

Verified cddaf028

github : update pr template

ggerganov committed 1 year ago

Verified c8a82194

Vulkan Shader Refactor, Memory Debugging Option (#7947)

0cc4m committed 1 year ago

Verified 7c7836d9

Add `cvector-generator` example (#7514)

ngxson committed 1 year ago

Verified 0c7b3595

[SYCL] remove global variables (#7710)

airMeng committed 1 year ago

Verified 7b2f4a7d

ci : fix macos x86 build (#7940)

olexiyb committed 1 year ago

Verified f8ec8877

CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (#7921)

JohannesGaessler committed 1 year ago

Verified 76d66ee0

metal : utilize max shared memory for mul_mat_id (#7935)

ggerganov committed 1 year ago

Verified 66ef1cee

llama-bench : fix RPC indication (#7936)

rgerganov committed 1 year ago

Verified e65bbf60

llama : more checks before assuming FIM tokens (#7644)

CISC committed 1 year ago

Verified 6fcd1331

convert : add Poro-34B-chat tokenizer support (#7713)

ezosa committed 1 year ago

Verified 41b9260f

rpc : fix ggml_backend_rpc_supports_buft() (#7918)

rgerganov committed 1 year ago

Verified 172c8256

readme : Remove outdated instructions from README.md (#7914) [no ci]

Galunid committed 1 year ago

Verified a55eb1bf

move BLAS to a separate backend (#6210)

slaren committed 1 year ago

Verified f578b86b

`build`: rename main → llama-cli, server → llama-server, llava-cli → llama-llava-cli, etc... (#7809)

ochafik committed 1 year ago

Verified 1c641e6a

Older