ggerganov/llama.cpp

Pull Requests Commits

server : remove old commented code [no ci]

ggerganov committed 60 days ago

Verified b8b17327

server : avoid common_batch

ggerganov committed 60 days ago

Verified 8a23b4a5

perplexity : avoid common_batch

ggerganov committed 61 days ago

Verified 76fd7d6f

embedding : avoid common_batch

ggerganov committed 62 days ago

Verified 8b80d683

retrieval : avoid common_batch

ggerganov committed 62 days ago

Verified 6f54ee66

android : fix permission

ngxson committed 62 days ago

32c2c41d

swift : adapt to new API

ggerganov committed 62 days ago

Verified 96ca6e8d

android : adapt to new API

ggerganov committed 62 days ago

Verified b0db7fc2

Merge pull request #15 from ggml-org/xsn/private_batch_api

ngxson committed 62 days ago

Verified 23d74073

speculative : adapt to new llama API

ggerganov committed 62 days ago

Verified 7a3c178d

Merge branch 'master' into xsn/private_batch_api

ngxson committed 62 days ago

dc4bb642

context : always use non-causal attention for encoder graphs (#12447)

ggerganov committed 63 days ago

Verified 8551c44d

SYCL: using graphs is configurable by environment variable and compile option (#12371)

lslusarczyk committed 63 days ago

Verified 35cae5ba

server : fix warmup draft cache type (#12446)

ggerganov committed 63 days ago

Verified 810e0af3

cmake : fix PowerPC build (#12241)

mehendarkarprajwal committed 63 days ago

Verified eba92d64

ggml : add SVE support for q6_K_q8_K (#12361)

fj-y-saito committed 63 days ago

Verified d9a14523

Vulkan: Default to 1GB allocations instead of 4GB to avoid fragmentation and driver issues (#12434)

0cc4m committed 63 days ago

Verified fd123cfe

fixed compilation warnings in ggml-sycl (#12424)

lslusarczyk committed 63 days ago

Verified a53f7f7b

llama: Add support for RWKV v7 architecture (#12412)

MollySophia committed 63 days ago

Verified 7dfad387

docs : bring llama-cli conversation/template docs up-to-date (#12426)

CISC committed 63 days ago

Verified 60c90292

cuda : enable CUDA Graph on CUDA Toolkit < 12.x (#12394)

gaugarg-nv committed 63 days ago

Verified b1b132ef

ggml-vulkan: remove unused find_program(glslc) (#12416)

guusw committed 63 days ago

Verified 01e8f213

vulkan: Add N/2 and N/4 optimized paths in coopmat2 shader (#12312)

jeffbolznv committed 63 days ago

Verified 484a8ab5

vulkan: subgroup size tuning (#12087)

daniandtheweb committed 64 days ago

Verified cf2270e4

Apply suggestions from code review

ngxson committed 64 days ago

Verified eab5606d

Update examples/tts/tts.cpp

ngxson committed 64 days ago

Verified de788e07

vulkan: use fp32 in coopmat2 q4_k dequant function (#12309)

jeffbolznv committed 64 days ago

Verified f07690c9

vulkan: Pad N dimension of B matrix for coopmat2 perf, to avoid bounds checking (#12273)

jeffbolznv committed 64 days ago

Verified 891c6395

vulkan: Adjust coopmat2 tile sizes and selection heuristic (#12258)

jeffbolznv committed 64 days ago

Verified 2f21123c

cmake : enable building llama.cpp using system libggml (#12321)

ckastner committed 64 days ago

Verified 374101fd

Older