ggerganov/llama.cpp

Pull Requests Commits

examples : add compare-mlx

ggerganov committed 39 days ago

Verified d8c17629

sampling : optimize samplers by reusing bucket sort (#15665)

ggerganov committed 40 days ago

Verified e92d53b2

server : enable /slots by default and make it secure (#15630)

ggerganov committed 40 days ago

Verified 0d161f02

metal : fix checks for available FA kernels (#15700)

ggerganov committed 40 days ago

Verified 4efd5a83

llama : fix fattn reserve call n_seqs parameter (#15699)

slaren committed 40 days ago

Verified 27496622

llama : separate compute buffer reserve from fattn check (#15696)

slaren committed 40 days ago

Verified 9777032d

ci : explicitly set fa off or on (#15692)

CISC committed 40 days ago

Verified 7d3c9f2b

vulkan: handle large sizes for get_rows (#15686)

jeffbolznv committed 40 days ago

Verified bbbf5ecc

vulkan: mul_mat_id coopmat2 optimizations (#15546)

jeffbolznv committed 40 days ago

Verified c37052ab

vulkan : remove unused portability_enumeration_ext variable (#15679)

danbev committed 40 days ago

Verified 5c16b9c8

vulkan: Allow fallback to sysmem memory when vidmem is full (#15649)

jeffbolznv committed 40 days ago

Verified b97c9edc

vulkan: clamp matmul and FA results to the max finite value (#15652)

jeffbolznv committed 40 days ago

Verified 94e82c7e

ggml: update kleidiai to v1.13.0 (#15663)

chaxu01 committed 41 days ago

Verified 4d74393b

Update build.md to remove MSVC arm64 notes (#15684)

slaren committed 41 days ago

Verified dd892555

llama: use FA + max. GPU layers by default (#15434)

JohannesGaessler committed 41 days ago

Verified e81b8e4b

CUDA: use FP32 arithmetic for conv2d (#15683)

JohannesGaessler committed 41 days ago

Verified 38ad381f

vulkan: Skip syncing for prealloc_y when it is reused (#15544)

jeffbolznv committed 41 days ago

Verified 696fccf3

CANN: FIx compiler warnings (#15661)

noemotiovon committed 41 days ago

Verified ef476916

server : removed obsolete doc (#15670)

l29ah committed 41 days ago

Verified d82f6aa3

scripts: strip "AMD Instinct" from GPU name (#15668)

JohannesGaessler committed 42 days ago

Verified 3d16b29c

server : add documentation for `parallel_tool_calls` param (#15647)

ExtReMLapin committed 42 days ago

Verified 792b44f2

CUDA: fix bug in rms_norm fusion (#15660)

am17an committed 42 days ago

Verified 81017865

chat : Seed OSS thinking + tool call support (#15552)

pwilkin committed 42 days ago

Verified 60e5eee3

CUDA: fuse adds, fuse add with rms norm (#15631)

am17an committed 42 days ago

Verified 009b709d

nvidia nemotron nano v2 (nemotronh) (#15507)

gabe-l-hart committed 42 days ago

Verified e8d99dd0

fix: Compute the full sum in llama-eval-callback, not just the sum of printed values (#15637)

gabe-l-hart committed 43 days ago

Verified a8bca68f

CUDA: add conv2d (#15635)

mnehete32 committed 43 days ago

Verified c97dc093

ggml-cpu: fix invalid hsum build in debug s390x (#15634)

taronaeo committed 43 days ago

Verified 6c442f42

ggml : fix SSM_SCAN for n_groups > 1 (#15625)

compilade committed 43 days ago

Verified 73804145

kv-cache : fix find_slot to not search for continuous slot (#15638)

ggerganov committed 43 days ago

Verified c8d0d14e

Older