ggml-org/llama.cpp

Pull Requests Commits

args : add -kvu to llama-parallel

ggerganov committed 28 days ago

Verified 5da56dc1

metal : fix build

ggerganov committed 28 days ago

Verified f8feadb2

Merge pull request #4 from gaugarg-nv/minor_fixes

JohannesGaessler committed 28 days ago

Verified b12a5635

Apply suggestion from @JohannesGaessler

JohannesGaessler committed 28 days ago

Verified 9bb9d783

Fix the seg fault without NCCL

gaugarg-nv committed 28 days ago

10385e8f

2d tensor set/get support

JohannesGaessler committed 29 days ago

3fdd0b7a

move allocation workaround out of ggml-alloc.c

JohannesGaessler committed 29 days ago

76d94392

Remove shfl and AllReduce from backend interface

JohannesGaessler committed 29 days ago

4dc3d10e

GGML: HIP: add RCCL support

IMbackK committed 29 days ago

29c5327d

JohannesGaessler committed 29 days ago

8de41b5b

fix output pattern

JohannesGaessler committed 29 days ago

c5314444

re-use buffers + ggml contexts

JohannesGaessler committed 29 days ago

c9255634

unconditional peer access

JohannesGaessler committed 29 days ago

02325685

add support for 4/8 GPUs

JohannesGaessler committed 29 days ago

2ffa49de

partial Vulkan fix

JohannesGaessler committed 29 days ago

4b8aa266

support for GPT-OSS, Qwen 3 MoE

JohannesGaessler committed 29 days ago

ab69c58a

ggml: backend-agnostic tensor parallelism

JohannesGaessler committed 29 days ago

a0d9dd20

metal : extend l2_norm support for non-cont src0 (#19502)

ggerganov committed 29 days ago

Verified 9ab072eb

docs: ban AI for issues and discussions [no CI] (#19512)

JohannesGaessler committed 29 days ago

Verified ada90bf2

common : improve download error reporting (#19491)

angt committed 29 days ago

Verified 0c1f39a9

hexagon: Add ARGSORT, DIV, SQR, SQRT, SUM_ROWS, GEGLU (#19406)

max-krasnyansky committed 29 days ago

Verified 73cd5e1b

llama : correct typos 'occured' and 'occurences' (#19414)

thecaptain789 committed 29 days ago

Verified 8ee538ce

model : fix wavtokenizer embedding notions (#19479)

ggerganov committed 29 days ago

Verified 6d957078

ggml : extend bin bcast for permuted src1 (#19484)

ggerganov committed 29 days ago

Verified 89181c0b

metal : consolidate unary ops (#19490)

ggerganov committed 29 days ago

Verified ceaa89b7

llama : refactor sampling_info to use buffer_view template (#19368)

danbev committed 29 days ago

Verified 2cce9fdd

CUDA : Update CCCL-tag for 3.2 to final release from RC (#19486)

ORippler committed 30 days ago

Verified 612db618

[WebGPU] Plug memory leaks and free resources on shutdown (#19315)

nikhilJain17 committed 30 days ago

Verified 57487a64

models : support qwen3.5 series (#19468)

JJJYmmm committed 30 days ago

Verified fc0fe400

test: fix IMROPE perf test case (#19465)

ngxson committed 30 days ago

Verified 9a963527

Older