ngxson/llama.cpp

Pull Requests Commits
CUDA: fix build error from ambiguous __half conversions in conv2d (#15690)
qnixsynapse committed 1 year ago
Verified b66df9d9
CANN: Optimize MUL_MAT_ID (#15658)
hipudding committed 1 year ago
Verified b9382c38
CANN: fix RoPE cache issue on multi-device (#15629)
hipudding committed 1 year ago
Verified 3dc7397a
sampling : optimize samplers by reusing bucket sort (#15665)
ggerganov committed 1 year ago
Verified e92d53b2
server : enable /slots by default and make it secure (#15630)
ggerganov committed 1 year ago
Verified 0d161f02
metal : fix checks for available FA kernels (#15700)
ggerganov committed 1 year ago
Verified 4efd5a83
llama : fix fattn reserve call n_seqs parameter (#15699)
slaren committed 1 year ago
Verified 27496622
llama : separate compute buffer reserve from fattn check (#15696)
slaren committed 1 year ago
Verified 9777032d
ci : explicitly set fa off or on (#15692)
CISC committed 1 year ago
Verified 7d3c9f2b
vulkan: handle large sizes for get_rows (#15686)
jeffbolznv committed 1 year ago
Verified bbbf5ecc
vulkan: mul_mat_id coopmat2 optimizations (#15546)
jeffbolznv committed 1 year ago
Verified c37052ab
vulkan : remove unused portability_enumeration_ext variable (#15679)
danbev committed 1 year ago
Verified 5c16b9c8
vulkan: Allow fallback to sysmem memory when vidmem is full (#15649)
jeffbolznv committed 1 year ago
Verified b97c9edc
vulkan: clamp matmul and FA results to the max finite value (#15652)
jeffbolznv committed 1 year ago
Verified 94e82c7e
ggml: update kleidiai to v1.13.0 (#15663)
chaxu01 committed 1 year ago
Verified 4d74393b
Update build.md to remove MSVC arm64 notes (#15684)
slaren committed 1 year ago
Verified dd892555
llama: use FA + max. GPU layers by default (#15434)
JohannesGaessler committed 1 year ago
Verified e81b8e4b
CUDA: use FP32 arithmetic for conv2d (#15683)
JohannesGaessler committed 1 year ago
Verified 38ad381f
vulkan: Skip syncing for prealloc_y when it is reused (#15544)
jeffbolznv committed 1 year ago
Verified 696fccf3
CANN: FIx compiler warnings (#15661)
noemotiovon committed 1 year ago
Verified ef476916
server : removed obsolete doc (#15670)
l29ah committed 1 year ago
Verified d82f6aa3
scripts: strip "AMD Instinct" from GPU name (#15668)
JohannesGaessler committed 1 year ago
Verified 3d16b29c
server : add documentation for `parallel_tool_calls` param (#15647)
ExtReMLapin committed 1 year ago
Verified 792b44f2
CUDA: fix bug in rms_norm fusion (#15660)
am17an committed 1 year ago
Verified 81017865
chat : Seed OSS thinking + tool call support (#15552)
pwilkin committed 1 year ago
Verified 60e5eee3
CUDA: fuse adds, fuse add with rms norm (#15631)
am17an committed 1 year ago
Verified 009b709d
nvidia nemotron nano v2 (nemotronh) (#15507)
gabe-l-hart committed 1 year ago
Verified e8d99dd0
fix: Compute the full sum in llama-eval-callback, not just the sum of printed values (#15637)
gabe-l-hart committed 1 year ago
Verified a8bca68f
CUDA: add conv2d (#15635)
mnehete32 committed 1 year ago
Verified c97dc093
ggml-cpu: fix invalid hsum build in debug s390x (#15634)
taronaeo committed 1 year ago
Verified 6c442f42
Terms Of Use Privacy Policy Refund Policy Impressum