ggml-org/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-08-19 18:45:26 +08:00
Code Issues Packages Projects Releases Wiki Activity
10,505 Commits 712 Branches 7,121 Tags

Default Branch

master

ee4c505a4f · server: add dedup-cache-models preset option (#27346) · Updated 2026-08-19 17:04:26 +08:00
 ZIP  TAR.GZ

Branches

gg/fa-no-kq-pad-save

837b1b4563 · ggml : remove KQ mask padding · Updated 2025-09-28 23:10:17 +08:00    tqcq

3882
6
 ZIP  TAR.GZ
llama-pull

17ca6ed540 · Implement llama-pull tool · Updated 2025-09-21 00:25:21 +08:00    tqcq

3970
1
 ZIP  TAR.GZ
xsn/qwen3next_experiment

e83ef74733 · one less magic number · Updated 2025-09-20 13:58:36 +08:00    tqcq

3989
6
 ZIP  TAR.GZ
gg/metal-fuse-add-rms

652d303b32 · metal : fuse add + rms · Updated 2025-09-18 21:29:25 +08:00    tqcq

3987
1
 ZIP  TAR.GZ
gg/metal-mul-mv-opt-2

64c6dcbe6d · metal : make the NSG a function constant in mul_mv kernels · Updated 2025-09-18 16:31:59 +08:00    tqcq

3992
2
 ZIP  TAR.GZ
gg/metal-use-virtual-gpu-address

6045c5a263 · cont : put all buffers in the same virtual address space · Updated 2025-09-14 20:46:57 +08:00    tqcq

4028
2
 ZIP  TAR.GZ
gg/metal-async

3f62ee8bee · metal : back to a single queue per device · Updated 2025-09-09 22:06:46 +08:00    tqcq

4070
9
 ZIP  TAR.GZ
gg/metal-async-save-global-queue

3f62ee8bee · metal : back to a single queue per device · Updated 2025-09-09 22:06:46 +08:00    tqcq

4070
9
 ZIP  TAR.GZ
rewrite-llama-run-to-be-llama-server-based

7b717fb4b2 · Rewrite llama-run to use llama-server · Updated 2025-09-06 00:22:36 +08:00    tqcq

4107
1
 ZIP  TAR.GZ
gg/metal-f16

9f2636b7dc · wip · Updated 2025-09-01 16:17:56 +08:00    tqcq

4156
1
 ZIP  TAR.GZ
gg/encode-pad-equal

4317d5abf5 · wip · Updated 2025-08-28 18:55:21 +08:00    tqcq

4190
1
 ZIP  TAR.GZ
compilade/fix-ssm-scan-groups

dc2187d48d · ggml : fix SSM_SCAN for n_groups > 1 · Updated 2025-08-28 05:37:04 +08:00    tqcq

4195
1
 ZIP  TAR.GZ
compilade/fix-qp-iq1-problems

fb573f4440 · ggml-quants : avoid division by zero in make_q3_quants · Updated 2025-08-18 06:26:02 +08:00    tqcq

4310
2
 ZIP  TAR.GZ
gg/graph-prec

220860aa0c · graph : use F32 accumulators for gpt-oss · Updated 2025-08-14 21:08:31 +08:00    tqcq

4337
1
 ZIP  TAR.GZ
compilade/imatrix-mxfp4

d9b625edb6 · ggml-quants : handle imatrix for MXFP4 · Updated 2025-08-12 10:12:10 +08:00    tqcq

4363
1
 ZIP  TAR.GZ
compilade/gguf-py-mxfp4

2763dc8b53 · ggml-quants : handle zero amax for MXFP4 · Updated 2025-08-07 04:26:25 +08:00    tqcq

4401
2
 ZIP  TAR.GZ
0cc4m/vulkan-op-opt-step-sgd

2ec70c964b · tests: Fix OPT_STEP_SGD test-backend-ops · Updated 2025-08-05 12:57:14 +08:00    tqcq

4409
4
 ZIP  TAR.GZ
compilade/fix-output-overflow

145401c9e3 · context : fix logits size overflow for huge batches · Updated 2025-08-05 10:26:46 +08:00    tqcq

4408
2
 ZIP  TAR.GZ
compilade/imatrix-gguf-warning

342e7014db · imatrix : only warn about suffix when output format is unspecified · Updated 2025-08-05 03:12:27 +08:00    tqcq

4413
2
 ZIP  TAR.GZ
compilade/hybrid-kv_unified

e549515cb3 · memory : handle kv_unified for hybrid models · Updated 2025-08-03 12:45:47 +08:00    tqcq

4422
1
 ZIP  TAR.GZ