ggerganov/llama.cpp

Pull Requests Commits

meta : tmp patch

ggerganov committed 4 days ago

Verified b59c34a5

glm5-next : cont q_absorbed for -sm tensor with non-unified KV

am17an committed 4 days ago

Verified 53db9b38

meta: support GLM-5.3 next sm tensor

am17an committed 4 days ago

Verified e3db613f

Change ggml_backend_comm_get_unique_id_t and ggml_backend_comm_init_rank_t to more generic multi-process backend communicator interfaces

zyang-dev committed 4 days ago

Verified 076f34cc

reuse ggml_backend_cuda_comm_allreduce_nccl for per-rank communicators. Remove the helper function backend_comm_proc_address

zyang-dev committed 4 days ago

Verified df1d0641

Drop ggml_backend_comm_allreduce_rank_t and reuse ggml_backend_comm_allreduce_tensor for per-rank communicators

zyang-dev committed 4 days ago

Verified 5ae06c7c

Add a backend interface for collectives across ranks running in separate processes, implement it in the CUDA backend using NCCL, and use it in the RPC server when available.

zyang-dev committed 4 days ago

Verified b0fc98e5

ui: fix system message row stretching to full prompt height in WebKit (#31246)

allozaur committed 4 days ago

Verified 3ab37aed

server: add router device memory margin parameter for dynamic unloading (#21231)

0cc4m committed 4 days ago

Verified 85cc432e

common : --model or -hf is required (#31242)

julien-c committed 4 days ago

Verified 74a1fd3d

model: Add MiniCPM-OCR-Jev (#30980)

tc-mb committed 4 days ago

Verified 7eed6b76

metal : add GGML_OP_TOP_K_MASK (#31230)

ggerganov committed 4 days ago

Verified 27637851

glm5-next : QSA without kq_mask (+ DSA) (#31009)

sanmai committed 4 days ago

Verified d78f40d9

ci : use ccache depend mode for ROCm release builds (#31231)

sreeram-11 committed 4 days ago

Verified fae9ac56

ci : force python 3.12 on ubuntu 26.04 (#31222)

CISC committed 4 days ago

Verified a8276b70

ggml : relax scale to non-contiguous views (#31146)

ggerganov committed 4 days ago

Verified de5122d8

tests : fix multi-seq split replay and partial-state cases for trained models (#31216)

ggerganov committed 4 days ago

Verified 6666a733

llama : remove obsolete n_pad from memory instances (#31200)

ggerganov committed 4 days ago

Verified 7a2e5aba

ui: list org-less models first in the model selector (#31215)

allozaur committed 4 days ago

Verified 0d5ca600

opencl: refine mul mat id -1 index (#31170)

lhez committed 4 days ago

Verified d930fe3c

opencl: initial support for q3_k (#31172)

lhez committed 4 days ago

Verified a2b3e121

ui: Highlight reasoning toggle on hover and align it with model selector (#31212)

allozaur committed 4 days ago

Verified 823df7d0

ui: add conversation compaction (#27401)

ServeurpersoCom committed 4 days ago

Verified 1a97e387

ui: Open model details in a drawer without leaving the page (#31176)

allozaur committed 4 days ago

Verified bc262f33

ui: use shared confirmation dialog for conversation deletion (#31187)

allozaur committed 4 days ago

Verified f5892ca5

hexagon: add GGML_HEXAGON_VTCM_SIZE (#31221)

ebateni committed 4 days ago

Verified 7aca0d2e

llama : enable moe amax rescaling for laguna (#31147)

begcdn committed 4 days ago

Verified 961c376a

ci : add windows behavior check (#31191)

CISC committed 4 days ago

Verified 4244ad7a

opencl: write KV-cache views in set_tensor so a state restore is not dropped (#30471)

wanghqc committed 4 days ago

Verified 63b4b303

hexagon: MUL_MAT_ID row skipping optimizations and improvements (#31169)

max-krasnyansky committed 4 days ago

Verified caa8191d

Older