ggml-org/llama.cpp

Pull Requests Commits

ggml-cuda: fix unused parameter in non-HIP builds

Victor-Loos committed 13 days ago

624617ac

ggml-cuda: avoid ROCm_Host compute on HIP integrated GPUs

Victor-Loos committed 13 days ago

52ff274b

Add tensor name to JSON output

pwilkin committed 13 days ago

5efd7400

tentative Metal support

pwilkin committed 13 days ago

4a4d3510

Add missing unrolls

pwilkin committed 13 days ago

7bbb7a17

pwilkin committed 13 days ago

bd6ec8e9

Fix FATTN profiling

pwilkin committed 13 days ago

c6dec166

Converge implementation with export-graph-ops

pwilkin committed 13 days ago

f6919937

Add missing op parameters to the profiler; add support for test-backend-ops to run performance tests with exactly the tensor shapes from the run

pwilkin committed 13 days ago

81f8b35d

docs, pass copy details

pwilkin committed 13 days ago

085847cd

fix mul_mat_id stats, add throughput stat, add envvar trigger, add concurrent mode fix

pwilkin committed 13 days ago

b6e0078b

fix builds, integrate vulkan profiler, fix copy events, fix export

pwilkin committed 13 days ago

1608f866

Fix more missing backend stuff (and Python errors)

pwilkin committed 13 days ago

b00e3a53

add second dimension to reported tensors, fix Mac build, add missing initializer to all backends

pwilkin committed 13 days ago

834067f6

feat: cool profiler thingy

pwilkin committed 13 days ago

19028b55

ci : add test-llama-archs tensor split for Metal (#27598)

ggerganov committed 13 days ago

Verified 95b8e33e

contrib : recommend waiting for CI before merging (#27603)

nikwen committed 13 days ago

Verified a278dcef

server : add LLAMA_SERVER_SLOTS_N_DIFF (#27600)

ggerganov committed 13 days ago

Verified e8eed452

common : skip device_info loop if it's not going to be printed (#26692)

wolfpld committed 13 days ago

Verified ba8e0edd

DeepseekV4: fix rollback with multi-seq (#26756)

am17an committed 13 days ago

Verified b0539c43

[Tensor parallel] Fix meta tensor split state propagation (#27574)

gaugarg-nv committed 13 days ago

Verified d3371929

ui: Chat Conversation Tabbed navigation (#27263)

allozaur committed 13 days ago

Verified 8144f319

vendor : update subprocess.h (#27409)

cabelo committed 13 days ago

Verified 6657ded4

cuda : add POOL_1D support (#27573)

amankarki151 committed 13 days ago

Verified 29ea9412

common: json.h: fix clang lto (#27575)

ngxson committed 13 days ago

Verified 70adb1b4

vulkan : added the PAD_REFLECT_1D operation (#26586)

safiullah3915 committed 13 days ago

Verified 3f545bec

mtmd: use ggml_rope_set_offset (#27521)

ngxson committed 14 days ago

Verified b21e4de7

common: add json.h abstraction (#27511)

ngxson committed 14 days ago

Verified d9f918d2

fit: also take into account n_streams (#27496)

ngxson committed 14 days ago

Verified 2fb989b9

arg: remove -no-cnv from cli [no ci] (#27542)

ngxson committed 14 days ago

Verified 9fee29e9

Older