Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
126 commits
Select commit Hold shift + click to select a range
e0feb41
add native_arch build option (#2199)
ContradNamiseb Jul 6, 2025
24719a6
alternative default search configuration (#2200)
borg323 Jul 6, 2025
09307ec
make it possible to specify the default backend (#2188)
borg323 Jul 12, 2025
91fa747
set strict_uci_timing_ initially to true (#2203)
borg323 Jul 13, 2025
1cbd507
build support for onnx-trt (#2204)
borg323 Jul 14, 2025
7dab58a
add cuda 12 build (#2205)
borg323 Jul 14, 2025
5a85e2a
modifying 'check if node_limit was initialized' to not fail if node_l…
KarlKfoury Jul 15, 2025
7d901fd
fix memory limit overflow (#2058)
KarlKfoury Jul 15, 2025
c61a16d
checking if 'name' parameter was provided with 'setoption' before pro…
KarlKfoury Jul 15, 2025
97d4532
cuda blas backward compatibility (#1747)
borg323 Jul 16, 2025
790aaf4
prefetch cleanup (#1778)
borg323 Jul 16, 2025
edb2ea2
appveyor onnx build cleanup (#2206)
borg323 Jul 17, 2025
14b723c
custom setoption parsing (#2207)
borg323 Jul 17, 2025
e0683b7
better default and options to set onnx ir (#2209)
borg323 Jul 18, 2025
1afa0a9
update authors, changelog and version before branch (#2210)
borg323 Jul 18, 2025
668a3a8
Use system opencl.hpp if it is available (#2213)
Menkib64 Jul 21, 2025
6960622
try to fix mac binary uploads (#2212)
borg323 Jul 23, 2025
fc5cb59
add a build using the latest xcode version (#2217)
borg323 Jul 23, 2025
5061581
error out for rpe nets (#2218)
borg323 Jul 24, 2025
60f030f
Fix warnings in clang / macos compile process (#2216)
almaudoh Jul 25, 2025
27283f6
Optimize metal: input expansion (#2220)
almaudoh Jul 27, 2025
64dd204
add onnx model hash to trt cache prefix (#2214)
borg323 Jul 28, 2025
5f2ab5f
alternative fp16 conversions using _Float where supported (#2219)
borg323 Jul 28, 2025
d915e36
Optimize metal: policy mapping (#2221)
almaudoh Jul 28, 2025
39c5de0
Sycl for AMD build improvements (#2215)
Menkib64 Jul 28, 2025
bdb1884
Support new hipBLAS 3 API (#2222)
Menkib64 Jul 29, 2025
ed2a400
remove c++17 workarounds (#2223)
borg323 Jul 29, 2025
b8cc55e
Update `sycl` backend (#2228)
ContradNamiseb Aug 1, 2025
4071355
Extract V6TrainingData struct to standalone header for external proje…
Copilot Aug 6, 2025
f00c619
fix some icx warnings (#2229)
borg323 Aug 6, 2025
92895f9
try to simplify installation of onnx-trt required dlls (#2225)
borg323 Aug 6, 2025
b7e9385
add onnx build option and default paths (#2226)
borg323 Aug 6, 2025
528ff3e
Make WDL_ShowWDL parameter false by default (#2234)
Copilot Aug 7, 2025
6bd14fa
Fixes to problems found because of Search::GetVerboseStats crash (#2232)
Menkib64 Aug 7, 2025
57c70f0
Fix Metal backend bug with out of bounds vector access (#2233)
ShukantPal Aug 7, 2025
b08633f
Avoid sycl waits in forwardEval code paths (#2230)
Menkib64 Aug 7, 2025
75e4617
fix sycl builds on windows (#2236)
borg323 Aug 8, 2025
d9ab724
appveyor only build cudnn on tags (#2238)
borg323 Aug 8, 2025
f538345
appveyor run tests with onnx (#2239)
borg323 Aug 8, 2025
04aaf8f
update cuda/cudnn version warning (#2240)
borg323 Aug 10, 2025
65ac6a4
Refactor ProcessFile function in rescorer.cc to improve maintainabili…
Copilot Aug 10, 2025
a41f7af
fix test dependencies (#2241)
borg323 Aug 11, 2025
2fc5152
Add improved SYCL device FP16 capability detection and device selecti…
ContradNamiseb Aug 11, 2025
ed3eb37
rescorer interface (#2246)
borg323 Aug 12, 2025
6132711
update changelog and authors for rc2 (#2244)
borg323 Aug 12, 2025
0081709
yet another try to fix mac binary uploads (#2250)
borg323 Aug 13, 2025
1dfa9bb
Use rocm_agent_enumerator to detect AMD GPU core version (#2254)
Menkib64 Aug 20, 2025
5b2c805
Start Count at One for Windows Install Instructions (#2252)
MrBrain295 Aug 20, 2025
319b83a
fix for eigen dependency issue (#2251)
borg323 Aug 20, 2025
70b8af6
updated cuda networks for Cuda TK 13 (#2256)
jackL999 Aug 21, 2025
acb7b24
readme update (#2253)
borg323 Aug 21, 2025
61487b2
update changelog and authors for release (#2257)
borg323 Aug 21, 2025
914d7b1
Support of default value in our protobufs (#2247)
mooskagh Aug 22, 2025
eb37e23
Output input embedding format in describenet (#2259)
mooskagh Aug 23, 2025
c29a755
build onnx on linux ci (#2262)
borg323 Aug 25, 2025
92d5c1b
make d_ double for dag (#2265)
borg323 Aug 25, 2025
b4e98c1
Avoid incrementing shared pointer when accessing low node (#2266)
Menkib64 Aug 25, 2025
2ec568d
Optimize dag node debugging code (#2267)
Menkib64 Aug 28, 2025
e6e07f3
Fix system eigen detection (#2269)
Menkib64 Aug 28, 2025
2ddf1a8
Silence thread sanitizer warnings in search (#2270)
Menkib64 Aug 28, 2025
1b685ff
Add debug messages to troubleshoot UCI problems (#2273)
Menkib64 Aug 31, 2025
f3e8776
Concurrent task workers (#2271)
Menkib64 Sep 3, 2025
769193c
Refactor: Replace legacy count() and find idioms with contains() (#2278)
SunnyWar Sep 6, 2025
eecd241
Update build for C++17 minimum standard: CUDA std=c++17, remove C++14…
SunnyWar Sep 6, 2025
b27a0b8
Fix param order (#2282)
SunnyWar Sep 7, 2025
0359c4f
remove submodule and move proto files to new directory (#2286)
borg323 Sep 12, 2025
1863b63
warning fixes (#2285)
borg323 Sep 12, 2025
425fae4
guard against inf fp16 input in softmax kernel (#2276)
borg323 Sep 13, 2025
8a072c6
Dag lock contention improvements (#2272)
Menkib64 Sep 13, 2025
3b42796
Use system libabsl if available (#2248)
Menkib64 Sep 15, 2025
f781098
fix meson warnings (#2290)
borg323 Sep 17, 2025
f20db7e
set output format for packaged onnx models (#2289)
borg323 Sep 17, 2025
e104e52
improve transpositon table memory use calculation (#2295)
borg323 Sep 26, 2025
13474cc
Fix piece moves marked en passant (#2292)
Menkib64 Sep 26, 2025
01cc532
fix pgn reader enpassant captures for black (#2301)
borg323 Oct 1, 2025
491c715
Fix use-after-free when old Search is destroyed (#2293)
Menkib64 Oct 5, 2025
8e30812
do onnx wdl softmax on the cpu (#2294)
borg323 Oct 5, 2025
540fcd1
Initialized activation functions in blas backend (#2300)
Menkib64 Oct 5, 2025
d2b4fb4
update to latest pjrt_c_api.h (#2302)
borg323 Oct 5, 2025
96c0beb
Add garbage collection thread to dag (#2274)
Menkib64 Oct 5, 2025
a0bd20f
make the install-trt script install everything needed (#2306)
borg323 Oct 10, 2025
d4c0a68
correct onnx mlh final activation (#2305)
borg323 Oct 10, 2025
204788a
Avoid allocating extra computation in search (#2310)
Menkib64 Oct 10, 2025
e36d650
Add threads option to BLAS backend (#2309)
Menkib64 Oct 10, 2025
f5fb402
Remove Sycl dpct dependency (#2296)
Menkib64 Oct 10, 2025
a50aad3
Refactor dag time keeping (#2288)
Menkib64 Oct 10, 2025
e22e80a
In the description of 'input' and 'output' which place between 'Lc0 w…
hans-ekbrand Oct 12, 2025
0a7ae40
fix build with dnnl blas (#2317)
borg323 Oct 15, 2025
bd6461c
fix rescorer refactoring bug (#2298)
borg323 Oct 15, 2025
0f68fc2
use default circleci mac resource class (#2325)
borg323 Oct 25, 2025
212ab93
fix onnx conversions with no gating (#2311)
borg323 Oct 25, 2025
9a9bb07
add support for fp16 encoded weights (#2283)
borg323 Oct 25, 2025
e6ecd4f
Fix Eigen math function build error. (#2316)
Menkib64 Oct 25, 2025
2599641
alternative mish approximation for onnx (#2326)
borg323 Oct 30, 2025
e9bd1c5
Make StrictTiming default and allow check for isready (#2327)
Menkib64 Oct 30, 2025
a37d608
attention policy head input may already be in nhwc format (#2334)
borg323 Nov 1, 2025
8195b0c
Support network evaluations per second in uci info (#2329)
Menkib64 Nov 1, 2025
4b16a98
Nvcc from cudnn path and debug symbols (#2335)
Menkib64 Nov 1, 2025
aacdc7f
Update Python bindings to work with new Move representation (#2299)
shermansiu Nov 3, 2025
3e49976
Onnx memory management and onnx-trt optimisation (#2307)
Menkib64 Nov 4, 2025
e2115ea
Print error location on Range violation. (#2330)
mooskagh Nov 7, 2025
bb616f9
calculate d as w and l in backend cpu softmax (#2337)
borg323 Nov 15, 2025
6de5d44
Add basic NVTX tracing support (#2331)
Menkib64 Nov 15, 2025
74a710c
Add bf16 support to onnx-trt (#2344)
Menkib64 Nov 15, 2025
26575ac
update mac circleci configuration (#2342)
borg323 Nov 18, 2025
d83051f
cleaner nvcc sharing for cuda and onnx in meson.build (#2341)
borg323 Nov 18, 2025
8417bc4
onnx full bf16 support requires opset 22 (#2345)
borg323 Nov 18, 2025
2537586
Cuda backend graph support (#2333)
Menkib64 Nov 18, 2025
dbfed8f
fix onnx locking (#2352)
borg323 Nov 22, 2025
67fa9e1
Backendbench threads and more statistics (#2351)
Menkib64 Nov 22, 2025
76b8925
Silence warning about modifying protected member. (#2347)
Menkib64 Nov 22, 2025
3f37760
update fp16 conversions (#2340)
borg323 Nov 30, 2025
9e715d2
Add coefficient of variation to backendbench (#2353)
Menkib64 Nov 30, 2025
310de9c
Add FLOAT32 encoding support for full precision weights (#2358)
Copilot Nov 30, 2025
a1b1197
Call cudaSetDevice only when GPU changes (#2359)
Menkib64 Nov 30, 2025
b600668
Demux refactoring(#2339)
Menkib64 Nov 30, 2025
9f35ac0
Add onnx-migraphx (#2363)
Menkib64 Dec 12, 2025
3bf83d5
onnx build cleanup (#2364)
borg323 Dec 14, 2025
acb4d47
onnx layernorm updates (#2355)
borg323 Dec 14, 2025
c3c3367
onnx alt mish in native float format (#2356)
borg323 Dec 14, 2025
6777ad2
cutlass fused multihead attention (#1976)
almaudoh Dec 20, 2025
9aba44a
Silence LTO warning about too large allocation (#2368)
Menkib64 Dec 20, 2025
be0079a
build cutlass as a subproject (#2369)
borg323 Dec 23, 2025
e482a65
move the onnx backend code in a directory (#2366)
borg323 Dec 23, 2025
711b7c0
Use strongly typed onnx-trt graphs (#2370)
Menkib64 Dec 30, 2025
7f572ae
Fix onnx-trt optimize option type (#2372)
Menkib64 Dec 30, 2025
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
64 changes: 41 additions & 23 deletions .circleci/config.yml
Original file line number Diff line number Diff line change
Expand Up @@ -9,24 +9,26 @@ jobs:
name: Install build tools
command: |
apt-get update
apt-get -y install git python3-pip gcc-10 g++-10 clang-12 zlib1g zlib1g-dev
apt-get -y install git python3-pip gcc-10 g++-10 clang-12 zlib1g zlib1g-dev wget
pip3 install meson==0.63
pip3 install ninja
- run:
name: "Pull Submodules"
command: git submodule update --init
name: Install onnxruntime
command: |
wget https://github.com/microsoft/onnxruntime/releases/download/v1.22.0/onnxruntime-linux-x64-1.22.0.tgz -P /tmp
tar xzf /tmp/onnxruntime-linux-x64-1.22.0.tgz -C /tmp
- run:
name: Meson GCC
environment:
CC: gcc-10
CXX: g++-10
command: meson build-gcc -Dgtest=false
command: meson build-gcc -Dgtest=false -Donnx_include=/tmp/onnxruntime-linux-x64-1.22.0/include -Donnx_libdir=/tmp/onnxruntime-linux-x64-1.22.0/lib
- run:
name: Meson Clang
environment:
CC: clang-12
CXX: clang++-12
command: meson build-clang -Dgtest=false -Db_lto=false
command: meson build-clang -Dgtest=false -Db_lto=false -Donnx_include=/tmp/onnxruntime-linux-x64-1.22.0/include -Donnx_libdir=/tmp/onnxruntime-linux-x64-1.22.0/lib
- run:
name: Build GCC
command: |
Expand All @@ -39,13 +41,9 @@ jobs:
ninja -j 4
"mac":
macos:
xcode: 14.1.0
resource_class: macos.m1.medium.gen1
xcode: 14.3.1
steps:
- checkout
- run:
name: "Pull Submodules"
command: git submodule update --init
- run:
name: Install build tools
command: |
Expand All @@ -71,43 +69,63 @@ jobs:
command: lipo -create -o /tmp/lc0 build/lc0 build-arm/lc0
- store_artifacts:
path: /tmp/lc0
destination: lc0-macos_12.6.1
destination: lc0-macos_13.2.1
- run:
name: Verify Workspace
name: Prepare Workspace
command: |
mv /tmp/lc0 /tmp/lc0-macos_12.6.1
ls -lah /tmp
mkdir -p workspace
mv /tmp/lc0 workspace
- persist_to_workspace:
root: /tmp
root: workspace
paths:
- lc0-macos_12.6.1

- lc0
"mac latest":
macos:
xcode: 26.1.0
steps:
- checkout
- run:
name: Install build tools
command: |
pip3 install meson
pip3 install ninja
- run:
name: Build lc0 arm
command: |
meson build-arm --buildtype=release -Dgtest=false -Dopencl=false
cd build-arm
ninja
"upload-github-release":
macos:
xcode: 14.1.0
xcode: 14.3.1
steps:
- attach_workspace:
at: /tmp
at: /tmp/workspace
- run:
name: Install GitHub CLI
command: brew install gh
- run:
name: Verify Workspace
command: |
ls -lah /tmp
ls -lah /tmp/workspace
- run:
name: Upload to GitHub Release
command: |
mv /tmp/workspace/lc0 /tmp/lc0-$CIRCLE_TAG-macos_13.2.1
gh release upload \
"$CIRCLE_TAG" \
/tmp/lc0-macos_12.6.1 \
--clobber
/tmp/lc0-$CIRCLE_TAG-macos_13.2.1 \
--clobber --repo LeelaChessZero/lc0
workflows:
version: 2
builds:
jobs:
- build
- "mac"
- "mac":
filters:
tags:
only: /v[0-9]+(\.[0-9]+)*(\-.+)?/
- "mac latest"
- "upload-github-release":
requires:
- "mac"
Expand Down
3 changes: 0 additions & 3 deletions .gitmodules
Original file line number Diff line number Diff line change
@@ -1,3 +0,0 @@
[submodule "libs/lczero-common"]
path = libs/lczero-common
url = https://github.com/LeelaChessZero/lczero-common.git
18 changes: 17 additions & 1 deletion AUTHORS
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@ almaudoh
Aloril
Andrew Grant
Andy Olsen
Aniebiet Udoh
Ankan
Ankan Banerjee
Anson Hu
Expand All @@ -19,9 +20,11 @@ Boštjan Mejak
Brandon Lin
Brett Holman
Carlo Wood
Chin-Chang Yang
cn4750
Cong
Contrad Namiseb (Bonan)
Copilot (bot)
cwbriscoe
danegraphics
Daniel Monroe
Expand All @@ -46,8 +49,10 @@ Francis Li
Francois
Francois Pays
François Pays
Gabe
Ganesh Krishnan
GBeauregard
Gergely Fülöp
Gian-Carlo Pascutto
gmorenz
Google LLC
Expand All @@ -56,11 +61,16 @@ Hace
Hans Ekbrand
Henrik Forstén
Ikko Eltociear Ashimine
Jack L
Jack Thomson
James Horsfall Thomas
jamie
jjoshua2
John Newlin
john-sp
Julian-Dominik Helmsen
Karl Kfoury
Kathleen Mcgrievy
kiilas
Kip Hamiltons
Kovax
Expand All @@ -73,6 +83,7 @@ Martin
Martin Senft
masterkni6
masterkni666
Menkib
Mike Roberts
Naphthalin
nathan-lc0
Expand All @@ -85,13 +96,17 @@ Pan
patrik-ha
PaulJeFi
Pratik Dixit
psykose
QxC4eva
Rafal Bielski
Raj
Reece H. Dunn
Ron Wolf
Sami Kiminki
Sherman Siu
Shreyas Kapur
shtayerc
Shukant Pal
Simon
slash
students
Expand All @@ -108,6 +123,7 @@ Valentin
Valeriy Huz
Victor Popovici
Videodr0me
Viet-Anh Tran
Viren6
Yan Zhang
zz4032
zz4032
Loading