Commits

Commits on May 5, 2025

Update CMakeLists.txt for tools/mtmd
JamePeng
committed
742cd08
Copy full SHA for 742cd08
Update Submodule vendor/llama.cpp 8ae5ebc..b34c859
JamePeng
committed
f1b41ae
Copy full SHA for f1b41ae

Commits on May 4, 2025

Update LLAVA_API code in llava_cpp.py
JamePeng
committed
939ca17
Copy full SHA for 939ca17
Update CMakeList.txt
JamePeng
committed
c5a8284
Copy full SHA for c5a8284
Update Submodule vendor/llama.cpp e1e8e09..8ae5ebc
JamePeng
committed
2e89e3f
Copy full SHA for 2e89e3f

Commits on May 3, 2025

Split cu124 and cu126 build-wheels-linux.yml:
Show description for 623f7ed
JamePeng
committed
623f7ed
Copy full SHA for 623f7ed
Try to enable LLAVA_BUILD in build-wheels-cuda-linux.yml
JamePeng
committed
c108044
Copy full SHA for c108044

Commits on May 2, 2025

Upload build-wheels-cuda-linux.yml
JamePeng
committed
786d15c
Copy full SHA for 786d15c

Commits on May 1, 2025

Update build-wheels-cuda-win.yml: use astral-sh/setup-uv@v6 for setup-uv option
JamePeng
committed
b0ea709
Copy full SHA for b0ea709
Sync llava_cpp code: Update clip.h function API
JamePeng
committed
0d2c985
Copy full SHA for 0d2c985
Update Submodule vendor/llama.cpp 3e168be..e1e8e09:
JamePeng
committed
faa7e05
Copy full SHA for faa7e05

Commits on Apr 30, 2025

Sync sampling : when top-k <= 0 -> noop (#13173)
JamePeng
committed
11e4438
Copy full SHA for 11e4438
Update Submodule vendor/llama.cpp 5630406..3e168be
JamePeng
committed
31725a2
Copy full SHA for 31725a2

Commits on Apr 23, 2025

Sync mtmd : Support Pixtral 12B (#13065)
JamePeng
committed
8b2e76e
Copy full SHA for 8b2e76e
Update Submodule vendor/llama.cpp 307bfa2..5630406
JamePeng
committed
d00e168
Copy full SHA for d00e168

Commits on Apr 14, 2025

Sync quantize: Handle user-defined quantization levels for additional tensors (#12511)
JamePeng
committed
c6b54d8
Copy full SHA for c6b54d8
Update Submodule vendor/llama.cpp 1d343b4..307bfa2
JamePeng
committed
56e7783
Copy full SHA for 56e7783

Commits on Apr 8, 2025

add -DLLAMA_CURL=OFF cmake arg avoid the the error "Could NOT find CURL"
JamePeng
committed
a505718
Copy full SHA for a505718
Sync llama : Support llama 4 text-only
JamePeng
committed
d390717
Copy full SHA for d390717
Update Submodule vendor/llama.cpp c262bed..8ca6e1c
JamePeng
committed
46d700f
Copy full SHA for 46d700f

Commits on Apr 4, 2025

Update llama : add option to override model tensor buffers
JamePeng
committed
20c1b27
Copy full SHA for 20c1b27
Update Submodule vendor/llama.cpp c80a775..c262bed
JamePeng
committed
fc9ca43
Copy full SHA for fc9ca43

Commits on Mar 31, 2025

Sync llama-vocab code 20250401
JamePeng
committed
7f4d94e
Copy full SHA for 7f4d94e
Update Submodule vendor/llama.cpp b3298fa..c80a775
JamePeng
committed
83e53f0
Copy full SHA for 83e53f0

Commits on Mar 26, 2025

Sync llama-vocab : add SuperBPE pre-tokenizer
JamePeng
committed
087566d
Copy full SHA for 087566d
Update Submodule vendor/llama.cpp e046430..b3298fa
JamePeng
committed
94f4358
Copy full SHA for 94f4358

Commits on Mar 21, 2025

update Submodule vendor/llama.cpp 8551c44..e046430
JamePeng
committed
cdc54d1
Copy full SHA for cdc54d1

Commits on Mar 18, 2025

Build Wheels (CUDA) for Windows workflow append build date tag
JamePeng
committed
b750a5f
Copy full SHA for b750a5f
Sync llama.cpp 20250319
JamePeng
committed
a5d6028
Copy full SHA for a5d6028
update Submodule vendor/llama.cpp 7841fc7..8551c44
JamePeng
committed
8491d8d
Copy full SHA for 8491d8d

Commits on Mar 13, 2025

class LlamaSampler: append add_xtc(), add_top_n_sigma() and add_dry()
Show description for c21dc41
JamePeng
committed
c21dc41
Copy full SHA for c21dc41
Add new Build Wheels (CUDA) for Windows workflow
Show description for 37ae873
JamePeng
committed
37ae873
Copy full SHA for 37ae873

Commits on Mar 12, 2025

chore: Bump version
abetlen
committed
37eb5f0
Copy full SHA for 37eb5f0
feat: Update llama.cpp
abetlen
committed
e232fae
Copy full SHA for e232fae
feat: Update llama.cpp
abetlen
committed
344c106
Copy full SHA for 344c106