ngxson/llama.cpp

Pull Requests Commits

Merge branch 'master' into xsn/n_cmpl_sync_barrier

ngxson committed 142 days ago

64b48eb4

context : reserve new scheduler when graph topology changes (#18547)

ggerganov committed 142 days ago

Verified 39173bca

CUDA: fix allignment on register spill for FA (#18815)

JohannesGaessler committed 142 days ago

Verified 5c662d21

correct return type of launch_slots_with_parent_task

ngxson committed 142 days ago

8b5474a0

Merge branch 'master' into xsn/n_cmpl_sync_barrier

ngxson committed 142 days ago

d5505b16

ggml-cpu: optimize ggml_vec_dot_bf16 for Power9 (#18837)

shalinib-ibm committed 142 days ago

Verified 8cc0ba95

lora: make sure model keep track of associated adapters (#18490)

ngxson committed 142 days ago

Verified a7e6ddb8

model-loader : support bool array sliding window pattern (#18850)

CISC committed 142 days ago

Verified 2a131801

tests : download models only when running ctest (#18843)

angt committed 142 days ago

Verified ec997b4f

hexagon: support for OP_CPY, host buffers now optional, hvx-utils refactoring and optimizations (#18822)

max-krasnyansky committed 142 days ago

Verified cff777f2

CUDA: Factor out and re-use `block_reduce` function (#18785)

ORippler committed 143 days ago

Verified 36f01324

ngxson committed 143 days ago

79c1967e

Restore clip's cb() to its rightful glory - extract common debugging elements in llama (#17914)

pwilkin committed 143 days ago

Verified d98b5481

model : clean up and fix EXAONE-MoE configuration (#18840)

nuxlear committed 143 days ago

Verified 8fb71755

simple approach

ngxson committed 143 days ago

ba86ad96

refactor : remove libcurl, use OpenSSL when available (#18828)

angt committed 143 days ago

Verified 516a4ca9

vulkan: Check maxStorageBufferRange in supports_op (#18709)

jeffbolznv committed 143 days ago

Verified 3e4bb296

llama-model: fix unfortunate typo (#18832)

am17an committed 143 days ago

Verified 47f96124

CUDA : fix typo in clang pragma comment [no ci] (#18830)

danbev committed 143 days ago

Verified 01cbdfd7

vulkan: work around Intel fp16 bug in mmq (#18814)

0cc4m committed 143 days ago

Verified 635ef78e

ggml-metal: do not copy headers for embedded, use current binary dir for embedded (#18705)

DaAwesomeP committed 143 days ago

Verified 7d587e55

mmap: add Haiku support by skipping RLIMIT_MEMLOCK check (#18819)

danielbenjaminsson committed 143 days ago

Verified d34aa071

ci, tests : use cmake to download models and remove libcurl dependency (#18791)

angt committed 143 days ago

Verified f709c7a3

llama : print_info alignment fix (#18708)

ddh0 committed 144 days ago

Verified 6e36299b

model : add EXAONE MoE (#18543)

nuxlear committed 144 days ago

Verified 60591f01

vocab : fix attribute overrides for harmony (#18806)

ggerganov committed 144 days ago

Verified e4832e3a

llama-mmap: fix direct-io loading fallback EOF exception (#18801)

0cc4m committed 144 days ago

Verified 960e5e3b

improve the logic

ngxson committed 144 days ago

d6b0d239

add pop_deferred_task with specific ID version

ngxson committed 144 days ago

da6e2bac

more debug messages

ngxson committed 144 days ago

f0349e40

Older