Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
358 commits
Select commit Hold shift + click to select a range
4ebb5cc
Merge pull request #4027 from codefuturedalao/master
wangzhaode Dec 22, 2025
edf9165
Merge pull request #4076 from jxt1234/feature/smallmodel_opt
wangzhaode Dec 22, 2025
afd359a
Merge branch feature/metal_backgroup_issue into master
wangzhaode Dec 23, 2025
731c263
Merge branch feature/fix_sync into master
wangzhaode Dec 23, 2025
38681b5
Merge branch feature/opencl_mmap_support into master
wangzhaode Dec 23, 2025
023dbea
refactor mdoel downloader
Juude Dec 23, 2025
fd07db1
MNN:Speed: Speed Up ReduceSum for inside >4
jxt1234 Dec 22, 2025
f012168
MNN:Speed: Reduce ConvolutionTiledExecutor's function enqueue cost
jxt1234 Dec 23, 2025
122fc81
Merge pull request #4081 from jxt1234/feature/reduce_conv_small_opt
wangzhaode Dec 24, 2025
e1acc84
Merge branch 'alibaba:master' into opt/rvv-c3-nv21
ihb2032 Dec 24, 2025
146aee1
Turn small group convolution to depthwise
jxt1234 Dec 23, 2025
f939b32
Project import generated by Copybara.
wangzhaode Dec 22, 2025
17486b5
Merge pull request #4067 from ihb2032/opt/rvv-pixel-conv
wangzhaode Dec 22, 2025
05183af
Merge pull request #4053 from ihb2032/opt/rvv-resize-functions
wangzhaode Dec 22, 2025
2f4c669
Merge pull request #4050 from ihb2032/opt/rvv-top1
wangzhaode Dec 22, 2025
7337385
Merge pull request #4044 from ihb2032/opt/rvv-softmax-relu
wangzhaode Dec 22, 2025
4e47bbd
Merge pull request #4042 from ihb2032/opt/rvv-conv-strassen
wangzhaode Dec 22, 2025
3f3ccf2
Merge pull request #4036 from ihb2032/opt/rvv-minmax-float
wangzhaode Dec 22, 2025
261ed0d
Merge pull request #4026 from ihb2032/opt/rvv-math-stride-ops
wangzhaode Dec 22, 2025
2f41cbe
Merge pull request #4023 from ihb2032/feature/rvv-transpose-functions
wangzhaode Dec 22, 2025
8928775
Merge pull request #4021 from ihb2032/feature/rvv-opt
wangzhaode Dec 22, 2025
941ed83
Merge pull request #4061 from zlaazlaa/fix_diffusion
wangzhaode Dec 22, 2025
b38d635
Merge pull request #3998 from bolun365/bolun365-patch-1
wangzhaode Dec 22, 2025
c9b6105
Merge pull request #4009 from HenryDen/default_opt
wangzhaode Dec 22, 2025
4aeed16
Merge branch feature/add_4th_groupchat into master
wangzhaode Dec 22, 2025
33fadc6
Merge pull request #4027 from codefuturedalao/master
wangzhaode Dec 22, 2025
a274567
Merge pull request #4076 from jxt1234/feature/smallmodel_opt
wangzhaode Dec 22, 2025
77643bc
Merge branch feature/metal_backgroup_issue into master
wangzhaode Dec 23, 2025
3edf873
Merge branch feature/fix_sync into master
wangzhaode Dec 23, 2025
f15a3d5
Project import generated by Copybara.
wangzhaode Dec 22, 2025
c5816f6
Merge pull request #4067 from ihb2032/opt/rvv-pixel-conv
wangzhaode Dec 22, 2025
22825af
Merge pull request #4053 from ihb2032/opt/rvv-resize-functions
wangzhaode Dec 22, 2025
b6cb9cf
Merge pull request #4050 from ihb2032/opt/rvv-top1
wangzhaode Dec 22, 2025
97363cb
Merge pull request #4044 from ihb2032/opt/rvv-softmax-relu
wangzhaode Dec 22, 2025
92d7832
Merge pull request #4042 from ihb2032/opt/rvv-conv-strassen
wangzhaode Dec 22, 2025
d484126
Merge pull request #4036 from ihb2032/opt/rvv-minmax-float
wangzhaode Dec 22, 2025
0994036
Merge pull request #4026 from ihb2032/opt/rvv-math-stride-ops
wangzhaode Dec 22, 2025
99184e3
Merge pull request #4023 from ihb2032/feature/rvv-transpose-functions
wangzhaode Dec 22, 2025
0cedcc4
Merge pull request #4021 from ihb2032/feature/rvv-opt
wangzhaode Dec 22, 2025
1e18ce0
Merge pull request #4061 from zlaazlaa/fix_diffusion
wangzhaode Dec 22, 2025
f20abc1
Merge pull request #3998 from bolun365/bolun365-patch-1
wangzhaode Dec 22, 2025
6fad876
Merge pull request #4009 from HenryDen/default_opt
wangzhaode Dec 22, 2025
1c88686
Merge branch feature/add_4th_groupchat into master
wangzhaode Dec 22, 2025
5654e26
Project import generated by Copybara.
wangzhaode Dec 22, 2025
1a854ac
Merge pull request #4067 from ihb2032/opt/rvv-pixel-conv
wangzhaode Dec 22, 2025
b607a75
Merge pull request #4053 from ihb2032/opt/rvv-resize-functions
wangzhaode Dec 22, 2025
f14c7f5
Merge pull request #4050 from ihb2032/opt/rvv-top1
wangzhaode Dec 22, 2025
4baf966
Merge pull request #4044 from ihb2032/opt/rvv-softmax-relu
wangzhaode Dec 22, 2025
0a46d79
Merge pull request #4042 from ihb2032/opt/rvv-conv-strassen
wangzhaode Dec 22, 2025
4f23caf
Merge pull request #4036 from ihb2032/opt/rvv-minmax-float
wangzhaode Dec 22, 2025
38cc110
Merge pull request #4026 from ihb2032/opt/rvv-math-stride-ops
wangzhaode Dec 22, 2025
161c501
Merge pull request #4023 from ihb2032/feature/rvv-transpose-functions
wangzhaode Dec 22, 2025
6289a3c
Merge pull request #4021 from ihb2032/feature/rvv-opt
wangzhaode Dec 22, 2025
a21ba9c
Merge pull request #4061 from zlaazlaa/fix_diffusion
wangzhaode Dec 22, 2025
a2f6fe8
Merge pull request #3998 from bolun365/bolun365-patch-1
wangzhaode Dec 22, 2025
e0982b3
Merge pull request #4009 from HenryDen/default_opt
wangzhaode Dec 22, 2025
6ccc2ac
Merge branch feature/add_4th_groupchat into master
wangzhaode Dec 22, 2025
aff7c58
Merge pull request #4027 from codefuturedalao/master
wangzhaode Dec 22, 2025
36c176b
Merge pull request #4076 from jxt1234/feature/smallmodel_opt
wangzhaode Dec 22, 2025
f85098f
Merge branch feature/metal_backgroup_issue into master
wangzhaode Dec 23, 2025
b45edc4
Merge branch feature/fix_sync into master
wangzhaode Dec 23, 2025
a9513c6
Merge branch feature/opencl_mmap_support into master
wangzhaode Dec 23, 2025
522f447
Merge pull request #4081 from jxt1234/feature/reduce_conv_small_opt
wangzhaode Dec 24, 2025
357257d
Merge branch feature/pyllm_status into master
wangzhaode Dec 25, 2025
3fc62a5
Merge branch Vulkan_buffer_fp16 into master
wangzhaode Dec 25, 2025
1a7ca3e
Merge branch feature/fix_loadtime into master
wangzhaode Dec 26, 2025
0b78f3d
Merge pull request #4017 from JunKnows/fix/opencl_depthwiseDeconv_and…
wangzhaode Dec 26, 2025
f99304e
Project import generated by Copybara.
wangzhaode Dec 26, 2025
a38df4e
Merge pull request #4082 from jxt1234/feature/smallconvolutionwithgro…
wangzhaode Dec 26, 2025
07bda85
Merge pull request #4079 from ihb2032/opt/rvv-c3-nv21
wangzhaode Dec 26, 2025
8dadab2
Merge pull request #4082 from jxt1234/feature/smallconvolutionwithgro…
wangzhaode Dec 26, 2025
c6ea9c4
Merge pull request #4079 from ihb2032/opt/rvv-c3-nv21
wangzhaode Dec 26, 2025
a280561
Merge branch feature/metal_fp32_bugfix into master
wangzhaode Dec 26, 2025
fa9c330
fix(rvv): add macro protection and system headers for cross-platform …
Dec 27, 2025
ca3dcb7
some typo corrections
jules-ai Dec 29, 2025
2fcbbd2
Merge branch feature/bugfix_axis into master
wangzhaode Dec 29, 2025
50233b8
Merge pull request #4088 from jules-ai/jules_fix_typo
wangzhaode Dec 30, 2025
6375755
Merge pull request #4085 from ihb2032/fix/rvv-macro-and-headers
wangzhaode Dec 30, 2025
bddc28f
MNN:Bugfix: Fix bug for CPUAttention with input size == 3
jxt1234 Dec 25, 2025
76b1e81
Converter:Bugfix: Fix bug for unary treat when input not has quant bu…
jxt1234 Dec 26, 2025
22517e7
Converter:Bugfix: Fix bug for don't copy des from onnx model
jxt1234 Dec 26, 2025
f3ddb62
Metal:Bugfix: Fix bug for not has mask or meta=nullptr
jxt1234 Dec 26, 2025
4605466
refactor download module to a common framework
Juude Dec 30, 2025
4add281
Merge pull request #4082 from jxt1234/feature/smallconvolutionwithgro…
wangzhaode Dec 26, 2025
2703cbe
Merge pull request #4079 from ihb2032/opt/rvv-c3-nv21
wangzhaode Dec 26, 2025
78c7471
Merge branch feature/metal_fp32_bugfix into master
wangzhaode Dec 26, 2025
ed50a24
Merge branch feature/bugfix_axis into master
wangzhaode Dec 29, 2025
fa72339
Merge pull request #4088 from jules-ai/jules_fix_typo
wangzhaode Dec 30, 2025
9b2b216
Merge pull request #4085 from ihb2032/fix/rvv-macro-and-headers
wangzhaode Dec 30, 2025
d3b2677
Merge branch feature/opencl_rerank_bugfix into master
wangzhaode Dec 30, 2025
5af9acc
Merge pull request #4083 from jxt1234/feature/bugfix
wangzhaode Dec 30, 2025
71ded12
Merge pull request #4082 from jxt1234/feature/smallconvolutionwithgro…
wangzhaode Dec 26, 2025
fb6048c
Merge pull request #4079 from ihb2032/opt/rvv-c3-nv21
wangzhaode Dec 26, 2025
dc1ba79
Merge branch feature/metal_fp32_bugfix into master
wangzhaode Dec 26, 2025
e251baf
Merge branch feature/bugfix_axis into master
wangzhaode Dec 29, 2025
7a0b6cf
Merge pull request #4088 from jules-ai/jules_fix_typo
wangzhaode Dec 30, 2025
b9b6416
Merge pull request #4085 from ihb2032/fix/rvv-macro-and-headers
wangzhaode Dec 30, 2025
f4ad319
Merge branch feature/opencl_rerank_bugfix into master
wangzhaode Dec 30, 2025
b6b1a64
Merge pull request #4083 from jxt1234/feature/bugfix
wangzhaode Dec 30, 2025
c1dc131
Merge branch feature/bugfix_fastvlmexport into master
wangzhaode Dec 30, 2025
e2a0374
1. Fixed download bugs 2. added MNN version number information 3. and…
Juude Dec 31, 2025
a17ee33
Merge branch feature/llm_nightly_test into master
wangzhaode Jan 4, 2026
eac58ed
Merge branch feature/all-zero-quant into master
wangzhaode Jan 4, 2026
3e2c4d8
add some tests
Juude Jan 4, 2026
5c25520
fix download framework bugs
Juude Jan 4, 2026
3a02456
add sherpa android demo
Juude Jan 4, 2026
6d9df9b
Merge branch feature/bugfix-vl-model into master
wangzhaode Jan 5, 2026
315a2c4
Geometry:Speed: Don't padding for zero case
jxt1234 Jan 5, 2026
162445c
Merge branch feature/del_log into master
wangzhaode Jan 5, 2026
3557f6f
MNN:Speed: Optimize LayerNorm for arm
jxt1234 Jan 5, 2026
e6d8259
add batch test scripts
Juude Jan 5, 2026
6a2d09d
MNN:Refractor: Add sparse macro for core function
jxt1234 Jan 6, 2026
8ec1be4
MNN:Refractor: Remove unusefule code fp32-fp8
jxt1234 Jan 6, 2026
adb9f68
MNN:Refractor:Remove unuseful code for attention
jxt1234 Jan 6, 2026
a1218cc
OpenCL:Bugfix: Fix bug for opencl compile error when debug
jxt1234 Jan 6, 2026
d3e0588
OpenCL:Bugfix: Reduce parameter to determine whether should use kvcache
jxt1234 Jan 6, 2026
dcf87d1
Tools:Feature: Support Collect Op list from model and prue_mnn_ops fr…
jxt1234 Jan 6, 2026
7904b05
Merge pull request #4097 from jxt1234/feature/oclbugfix
wangzhaode Jan 6, 2026
1296705
Merge pull request #4098 from jxt1234/feature/support_reduce_byop
wangzhaode Jan 6, 2026
c1c8456
Merge pull request #4095 from jxt1234/feature/speed
wangzhaode Jan 6, 2026
3408e29
docs: fix typos and improve CLI help text
AliasJeff Jan 7, 2026
c3ff0d9
Merge remote-tracking branch 'remotes/origin/master' into tts/app
Juude Jan 7, 2026
c7a20df
Merge pull request #4082 from jxt1234/feature/smallconvolutionwithgro…
wangzhaode Dec 26, 2025
2fd705d
Merge pull request #4079 from ihb2032/opt/rvv-c3-nv21
wangzhaode Dec 26, 2025
ef2883e
Merge branch feature/metal_fp32_bugfix into master
wangzhaode Dec 26, 2025
6020562
Merge branch feature/bugfix_axis into master
wangzhaode Dec 29, 2025
2dd01b4
Merge pull request #4088 from jules-ai/jules_fix_typo
wangzhaode Dec 30, 2025
0764ec2
Merge pull request #4085 from ihb2032/fix/rvv-macro-and-headers
wangzhaode Dec 30, 2025
2e1b72a
Merge branch feature/opencl_rerank_bugfix into master
wangzhaode Dec 30, 2025
17c5275
Merge pull request #4083 from jxt1234/feature/bugfix
wangzhaode Dec 30, 2025
a49c2b9
Merge branch feature/bugfix_fastvlmexport into master
wangzhaode Dec 30, 2025
5e5c70a
Merge branch feature/llm_nightly_test into master
wangzhaode Jan 4, 2026
6b1cbd9
Merge branch feature/all-zero-quant into master
wangzhaode Jan 4, 2026
e0495d5
Merge branch feature/bugfix-vl-model into master
wangzhaode Jan 5, 2026
3089d4d
Merge branch feature/del_log into master
wangzhaode Jan 5, 2026
6262f35
Merge pull request #4097 from jxt1234/feature/oclbugfix
wangzhaode Jan 6, 2026
12d7623
Merge pull request #4098 from jxt1234/feature/support_reduce_byop
wangzhaode Jan 6, 2026
5ec6d4c
Merge pull request #4095 from jxt1234/feature/speed
wangzhaode Jan 6, 2026
bd271b8
Merge branch feature/bugfix-llmbench into master
wangzhaode Jan 7, 2026
9bd8302
Merge pull request #4101 from Juude/tts/app
wangzhaode Jan 7, 2026
b3325f3
Merge pull request #4082 from jxt1234/feature/smallconvolutionwithgro…
wangzhaode Dec 26, 2025
1c79d7d
Merge pull request #4079 from ihb2032/opt/rvv-c3-nv21
wangzhaode Dec 26, 2025
f3ee090
Merge branch feature/metal_fp32_bugfix into master
wangzhaode Dec 26, 2025
24e782e
Merge branch feature/bugfix_axis into master
wangzhaode Dec 29, 2025
69db3af
Merge pull request #4088 from jules-ai/jules_fix_typo
wangzhaode Dec 30, 2025
1ec6338
Merge pull request #4085 from ihb2032/fix/rvv-macro-and-headers
wangzhaode Dec 30, 2025
d320eac
Merge branch feature/opencl_rerank_bugfix into master
wangzhaode Dec 30, 2025
d9b8afc
Merge pull request #4083 from jxt1234/feature/bugfix
wangzhaode Dec 30, 2025
55bf3e1
Merge branch feature/bugfix_fastvlmexport into master
wangzhaode Dec 30, 2025
3071997
Merge branch feature/llm_nightly_test into master
wangzhaode Jan 4, 2026
ed301f8
Merge branch feature/all-zero-quant into master
wangzhaode Jan 4, 2026
690e24f
Merge branch feature/bugfix-vl-model into master
wangzhaode Jan 5, 2026
9aaa43b
Merge branch feature/del_log into master
wangzhaode Jan 5, 2026
ac4fccf
Merge pull request #4097 from jxt1234/feature/oclbugfix
wangzhaode Jan 6, 2026
96915f6
Merge pull request #4098 from jxt1234/feature/support_reduce_byop
wangzhaode Jan 6, 2026
b7d1619
Merge pull request #4095 from jxt1234/feature/speed
wangzhaode Jan 6, 2026
1df2688
Merge branch feature/bugfix-llmbench into master
wangzhaode Jan 7, 2026
7ee98b3
Merge pull request #4101 from Juude/tts/app
wangzhaode Jan 7, 2026
11cd83d
Merge branch feature/bugfix-sse-softmax into master
wangzhaode Jan 8, 2026
0a5ebb9
Merge branch feature/asymmetric-llm-ptq into master
wangzhaode Jan 8, 2026
c0d15bf
Merge branch feature/ci_add_vlm into master
wangzhaode Jan 8, 2026
c6a6301
cache `MNN_DEPS` to `MNN_LIBS`, `MNN_INCLUDES` to `MNN_INCLUDE_DIRS` …
rainyl Jan 10, 2026
60c0e01
Merge branch feature/Vulkan_time_profile_bugfix into master
wangzhaode Jan 12, 2026
7a68561
Merge branch feature/Vulkan_buffer_layernorm_bugfix into master
wangzhaode Jan 12, 2026
e0295ae
Merge branch feature/llm_bench_json into master
wangzhaode Jan 12, 2026
550a740
Merge branch feature/opencl_attention_bugfix into master
wangzhaode Jan 13, 2026
f8f228c
Merge branch feature/eagle_export_bugfix into master
wangzhaode Jan 13, 2026
80c1695
MNN:Speed: Optimize LayerNorm's accu and norm function
jxt1234 Jan 13, 2026
0f9348a
Merge branch feature/bugfix-doc-npu into master
wangzhaode Jan 14, 2026
72a9a05
fix download error, support multiple images input
Juude Jan 14, 2026
1938e5b
fix inifite chars in mnncli interactive chat mode
Juude Jan 14, 2026
52d5e90
Update mnncli build script for Linux support
Juude Jan 14, 2026
baaa38b
Merge pull request #4112 from Juude/tts/app
wangzhaode Jan 15, 2026
05f6954
Merge branch feature/opencl_softmax_bugfix into master
wangzhaode Jan 15, 2026
ad29157
Merge branch feature/llmexport_support_skipweight into master
wangzhaode Jan 15, 2026
53ff0ac
Converter:Feature: Add convpad fuse
jxt1234 Jan 16, 2026
4129244
Fix quantization batch default for fixed-shape models
LudovicoYIN Jan 17, 2026
b1e89f1
Converter:Bugfix: Fix bug for prelu in w / h
jxt1234 Jan 19, 2026
a53fae4
Merge pull request #4117 from LudovicoYIN/fix/quant-batch-default
wangzhaode Jan 20, 2026
af80558
Merge pull request #4119 from jxt1234/feature/prelubugfix
wangzhaode Jan 20, 2026
6c7acf8
Merge pull request #4115 from jxt1234/feature/convpadfuse
wangzhaode Jan 20, 2026
e04ca23
Merge pull request #4108 from jxt1234/feature/speeduplayernorm
wangzhaode Jan 20, 2026
170c76f
Merge pull request #4106 from rainyl/cache-mnn-deps-includes
wangzhaode Jan 20, 2026
adf31e6
Merge pull request #4100 from AliasJeff/doc/fix-doc-typos
wangzhaode Jan 20, 2026
3af5286
Project import generated by Copybara.
wangzhaode Jan 21, 2026
912235a
[CI:Typo] typo transformer README for copybara sync.
wangzhaode Jan 21, 2026
770d8b3
Merge branch feature/fix_ppl into master
wangzhaode Jan 21, 2026
d7db2f1
Merge branch feature/Vulkan_time_prfoile_refactor into master
wangzhaode Jan 22, 2026
4fb2bc0
Merge branch feature/Vulkan_buffer_indirect_mode into master
wangzhaode Jan 22, 2026
11ba2cd
Express:Bugfix: Fix bug for create const op with external erro
jxt1234 Jan 22, 2026
02ef424
MNN:Bugfix: Fix bug for CPUAttention quant may overflow for Q
jxt1234 Jan 22, 2026
25685a0
Merge pull request #4096 from jxt1234/feature/bugfix
wangzhaode Jan 22, 2026
71685a0
fix: sumParams4QKxV is being used without being initialized
EricMoin Jan 25, 2026
d4c861b
Converter:Bugfix: Fix bug for check model with setting value
jxt1234 Jan 26, 2026
f0a6ffb
Merge pull request #4129 from jxt1234/feature/bugfix
wangzhaode Jan 27, 2026
75a2f50
Merge pull request #4127 from EricMoin/master
wangzhaode Jan 27, 2026
6b631ec
Merge branch feature/fix_ppl into master
wangzhaode Jan 21, 2026
e8f9a5e
Merge branch feature/Vulkan_time_prfoile_refactor into master
wangzhaode Jan 22, 2026
014d054
Merge branch feature/Vulkan_buffer_indirect_mode into master
wangzhaode Jan 22, 2026
fd25010
Merge pull request #4096 from jxt1234/feature/bugfix
wangzhaode Jan 22, 2026
954e1c2
Merge pull request #4129 from jxt1234/feature/bugfix
wangzhaode Jan 27, 2026
fa67eac
Merge pull request #4127 from EricMoin/master
wangzhaode Jan 27, 2026
ac5ab73
Merge branch feature/convert_dumppass into master
wangzhaode Jan 27, 2026
510ac8f
[MNN:Sync] Manual sync to internal 91b0f760069d86273cc3fd8f7f90ee504a…
wangzhaode Jan 27, 2026
36294e5
Merge branch feature/open_llm_pyc into master
wangzhaode Jan 27, 2026
1ce6a78
Merge branch feature/fix_ppl into master
wangzhaode Jan 21, 2026
64603ce
Merge branch feature/Vulkan_time_prfoile_refactor into master
wangzhaode Jan 22, 2026
48acab8
Merge branch feature/Vulkan_buffer_indirect_mode into master
wangzhaode Jan 22, 2026
2728959
Merge pull request #4096 from jxt1234/feature/bugfix
wangzhaode Jan 22, 2026
3ad142b
Merge pull request #4129 from jxt1234/feature/bugfix
wangzhaode Jan 27, 2026
df02a6d
Merge pull request #4127 from EricMoin/master
wangzhaode Jan 27, 2026
2b2d0c1
Merge branch feature/convert_dumppass into master
wangzhaode Jan 27, 2026
6dd4f07
Merge branch feature/open_llm_pyc into master
wangzhaode Jan 27, 2026
ca69dff
Merge branch feature/metal_falshattn into master
wangzhaode Jan 29, 2026
d8bc4f2
[MNN:Sync] Manual sync to internal.
wangzhaode Jan 29, 2026
6bf46be
[MNN:Sync] Manual sync to internal.
wangzhaode Feb 2, 2026
61448bc
Merge branch feature/fix_ppl into master
wangzhaode Jan 21, 2026
83ef2b6
Merge branch feature/Vulkan_time_prfoile_refactor into master
wangzhaode Jan 22, 2026
10e435f
Merge branch feature/Vulkan_buffer_indirect_mode into master
wangzhaode Jan 22, 2026
e751292
Merge pull request #4096 from jxt1234/feature/bugfix
wangzhaode Jan 22, 2026
9d918e5
Merge pull request #4129 from jxt1234/feature/bugfix
wangzhaode Jan 27, 2026
cc118ff
Merge pull request #4127 from EricMoin/master
wangzhaode Jan 27, 2026
a5ec769
Merge branch feature/convert_dumppass into master
wangzhaode Jan 27, 2026
38a341f
Merge branch feature/open_llm_pyc into master
wangzhaode Jan 27, 2026
9ed6157
Merge branch feature/metal_falshattn into master
wangzhaode Jan 29, 2026
30a8116
[MNN:Feature] CPU attention speed(softMax reducing cache miss) and bu…
wangzhaode Jan 5, 2026
8e9b791
[LLM:Feature] LoRA model support clone LayerNorm from base.
wangzhaode Feb 2, 2026
0244b85
feat: 增加 sana diffusion, 重构代码
wangzhaode Feb 2, 2026
2c996ad
Converter:Bugfix: Fix bug for ConvBiasAdd may change output name
jxt1234 Feb 3, 2026
4075f18
[LLM:Feature] Support Fun-Audio-Chat-8B.
wangzhaode Feb 4, 2026
d483455
[Reranker Feature] Support Loop with multi-commands
wangzhaode Jan 29, 2026
a874b30
[MNN:Version] Update to 3.4.0.
wangzhaode Feb 7, 2026
c430d76
[MNN:CI] Upgrade macos-13 to macos-14 for pymnn release workflow
wangzhaode Feb 9, 2026
4eb5ee6
[feature:bugfix]bugfix for reranker_demo crash without load model
wangzhaode Feb 9, 2026
fd56bb4
fix: 修改sana resize阈值
wangzhaode Feb 10, 2026
30a500c
[VULKAN][BUFFER] Support coopMat in Vulkan Conv1x1.
wangzhaode Feb 11, 2026
1fdc7d3
[MNN:Feature] Support Qwen3.5.
wangzhaode Feb 9, 2026
5541aac
Merge branch feature/support_qwen3.5_27b into master
wangzhaode Feb 27, 2026
ff895fa
[MNN:Security] fix memory safety vulnerabilities in multiple expr ope…
wangzhaode Feb 27, 2026
e30ad2c
feat: sana android and app updates with scoped docs and qnn deps
Juude Mar 1, 2026
7f1197e
Sana Diffusion Style Transfer, Omni Audio Output, Video Input & More …
wangzhaode Mar 2, 2026
81560d0
[LLM:Bugfix] prevent out-of-bounds crash for multimodal models in llm…
wangzhaode Mar 3, 2026
14ad08c
解决benchmark crash、add tests
Juude Mar 3, 2026
6203235
Merge branch feature/bugfix_jsonmerge into master
wangzhaode Mar 4, 2026
38c2033
[Bugfix] metal weighti8i4conv2d op test error
wangzhaode Feb 10, 2026
227791e
Metal:Feature: Support Clone for MetalConvolutionDepthwise
jxt1234 Mar 4, 2026
7fcf52c
[CI:Feature] add llm pr review.
wangzhaode Mar 4, 2026
bc55256
[LLM:bugfix] llm add executor and use ExecutorScope control it
Qxinyu Mar 4, 2026
6b1db4c
[MNN:Version] Update to 3.4.1.
wangzhaode Mar 5, 2026
a1803f7
chore: merge upstream MNN 3.4.1
pruthvikar Mar 26, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
24 changes: 24 additions & 0 deletions .github/workflows/llm-pr-review.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,24 @@
name: LLM Code Review

on:
pull_request:
types: [opened, reopened, synchronize]

permissions:
contents: read
pull-requests: write

jobs:
review:
runs-on: ubuntu-latest
steps:
- name: Checkout Repo
uses: actions/checkout@v4

- name: LLM Code Review
uses: wangzhaode/MNNCodeReviewer@v1.0.0
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
OPENAI_API_ENDPOINT: https://maas-api.ai-yuanjing.com/openapi/compatible-mode/v1
MODEL: glm-5
6 changes: 3 additions & 3 deletions .github/workflows/pymnn_release.yml
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ jobs:
- { os: ubuntu-latest, arch: x86_64, build: 'cp*-manylinux*' }
- { os: ubuntu-24.04-arm, arch: aarch64, build: 'cp*-manylinux*' }
- { os: windows-latest, arch: AMD64, build: 'cp*' }
- { os: macos-13, arch: x86_64, build: 'cp*' }
- { os: macos-14, arch: x86_64, build: 'cp*' }
- { os: macos-14, arch: arm64, build: 'cp*' }

steps:
Expand All @@ -39,7 +39,7 @@ jobs:
run: python -m pip install pipx

- name: Build wheels
uses: pypa/cibuildwheel@v2.16.5
uses: pypa/cibuildwheel@v2.22.0
env:
CIBW_ARCHS_MACOS: ${{ matrix.arch }}
CIBW_ARCHS_LINUX: ${{ matrix.arch }}
Expand Down Expand Up @@ -69,6 +69,7 @@ jobs:
publish_wheels:
permissions:
contents: none
id-token: write
name: Upload
needs: [build_wheels]
runs-on: ubuntu-latest
Expand All @@ -86,5 +87,4 @@ jobs:

- uses: pypa/gh-action-pypi-publish@release/v1
with:
password: ${{ secrets.PYPI_API_TOKEN }}
skip_existing: true
17 changes: 16 additions & 1 deletion .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -375,9 +375,24 @@ datasets/*
source/backend/qnn/3rdParty/include
project/android/.cxx
pymnn/android/.cxx/
pymnn/android/.cxx/abi_configuration_5u53tc49.jsonz
apps/mnncli/.cursorrules
apps/mnncli/model_market_json_data.inc
#kledi
_deps
#aicoding
.cursor
.cursor

# llm model
transformers/llm/export/model/
apps/Android/.qoder/settings.json
apps/Android/MnnLlmChatOld

transformers/llm/export/tmp/

# iOS
apps/iOS/MNNLLMChat/Chat/
apps/iOS/MNNLLMChat/swift-transformers/
apps/iOS/MNNLLMChat/MNNLLMiOS/LocalModel/Qwen3-4B-MNN
apps/iOS/MNNLLMChat/MNNLLMiOS/LocalModel/Qwen3-0.6B-MNN
apps/iOS/MNNLLMChat/MNNLLMiOS/LocalModel/Qwen2.5-Omni-3B-MNN
13 changes: 12 additions & 1 deletion CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,7 @@ option(MNN_LOW_MEMORY "Build MNN support low memory for weight quant model." OFF
option(MNN_CPU_WEIGHT_DEQUANT_GEMM "Build MNN CPU weight dequant related gemm kernels." OFF)
option(MNN_BUILD_AUDIO "Build audio api in MNN." OFF)
option(MNN_SME2 "Use Arm sme2 instructions" ON)
option(MNN_METAL_TENSOR "Use Metal4 tensor instructions" ON)

if (MNN_BUILD_MINI)
set(MNN_SKIPBUILD_GEOMETRY ON CACHE BOOL "<docstring>" FORCE)
Expand Down Expand Up @@ -258,6 +259,7 @@ option(MNN_VULKAN "Enable Vulkan" OFF)
option(MNN_ARM82 "Enable ARMv8.2's FP16 Compute" ON)
option(MNN_SUPPORT_FP16_ARMV7 "Enable ARMv8.2's FP16 Compute for armv7 arch, may cause library not valid for 32 bit cpu" OFF)
option(MNN_KLEIDIAI "Enable KLEIDIAI" ON)
option(MNN_KLEIDIAI_DEFAULT_ON "Use KLEIDIAI kernels by default" OFF)
option(MNN_ONEDNN "Enable oneDNN" OFF)
option(MNN_AVX2 "Open AVX2 Compile for x86 if possible" ON)
option(MNN_AVX512 "Enable AVX512" OFF)
Expand All @@ -277,7 +279,7 @@ if (NOT MNN_CUDA OR NOT CMAKE_SYSTEM_NAME MATCHES "^Linux")
set(MNN_CUDA_PROFILE OFF)
endif()

if (NOT MNN_QNN)
if (NOT MNN_QNN)
set(MNN_QNN_ONLINE_FINALIZE OFF)
endif()

Expand Down Expand Up @@ -373,6 +375,9 @@ endif()
IF(MNN_DEBUG_MEMORY)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=address")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=address")

set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -fsanitize=address")
set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} -fsanitize=address")
endif()

set(MNN_DEPS "")
Expand Down Expand Up @@ -549,6 +554,7 @@ ENDIF()
IF(MNN_BUILD_DIFFUSION)
file(GLOB MNN_DIFFUSION_HDRS ${CMAKE_CURRENT_SOURCE_DIR}/transformers/diffusion/engine/include/diffusion/*)
list(APPEND MNN_EXTRA_HEADERS ${CMAKE_CURRENT_SOURCE_DIR}/transformers/diffusion/engine/include/diffusion/diffusion.hpp)
list(APPEND MNN_EXTRA_HEADERS ${CMAKE_CURRENT_SOURCE_DIR}/transformers/diffusion/engine/include/diffusion/sana_llm.hpp)
ENDIF()


Expand Down Expand Up @@ -936,6 +942,11 @@ if (NOT MNN_BUILD_SHARED_LIBS)
endif()
list(APPEND MNN_TARGETS MNN)
list(REMOVE_ITEM MNN_TARGETS MNN)

# Cache MNN_DEPS and MNN_INCLUDES for external projects
set(MNN_LIBS ${MNN_DEPS} CACHE INTERNAL "MNN targets")
set(MNN_INCLUDE_DIRS ${MNN_INCLUDES} CACHE INTERNAL "MNN include directories")

IF(MNN_BUILD_DEMO)
include(${CMAKE_CURRENT_LIST_DIR}/demo/exec/CMakeLists.txt)
ENDIF()
Expand Down
Empty file removed MNN.sln
Empty file.
34 changes: 22 additions & 12 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,13 +6,27 @@
[![日本語バージョン](https://img.shields.io/badge/Language-%E6%97%A5%E6%9C%AC%E8%AA%9E-green)](README_JP.md)
[![MNN Homepage](https://img.shields.io/badge/Homepage-Visit-green)](http://www.mnn.zone)

[![MNN Chat App](https://img.shields.io/badge/Apps-MNN_Chat-blue)](./apps/Android/MnnLlmChat/README.md)
[![TaoAvatar](https://img.shields.io/badge/Apps-MNN_TaoAvatar-blue)](./apps/Android/Mnn3dAvatar/README.md)

[![MNN Chat App](https://img.shields.io/badge/Apps-MNN_Chat-blue)](./apps/Android/MnnLlmChat/README.md)
[![TaoAvatar](https://img.shields.io/badge/Apps-MNN_TaoAvatar-blue)](./apps/Android/Mnn3dAvatar/README.md)
[![Sana](https://img.shields.io/badge/Apps-Sana_Image_Edit-blue)](./apps/sana/README.md)

## News 🔥
- [2026/03/05] Support Qwen3.5 Series.
<p align="center">
<img width="15%" alt="Icon" src="https://meta.alicdn.com/data/mnn/assets/qwen35_1.jpg" style="margin: 0 10px;">
<img width="15%" alt="Icon" src="https://meta.alicdn.com/data/mnn/assets/qwen35_2.jpg" style="margin: 0 10px;">
<img width="15%" alt="Icon" src="https://meta.alicdn.com/data/mnn/assets/qwen35_3.jpg" style="margin: 0 10px;">
</p>
- [2026/02/13] MNN-Sana-Edit-V2 is now available at [apps](./apps/sana/README.md), offering cartoon-style photo editing based on Sana.
<p align="center">
<img width="80%" alt="Icon" src="https://meta.alicdn.com/data/mnn/assets/sana_show_case.jpg" style="margin: 0 10px;">
</p>

<details>
<summary> History News </summary>

- [2025/10/16] Support Qwen3-VL Series.
- [2025/06/11] New App MNN TaoAvatar released, you can talk with 3DAvatar offline with LLM, ASR, TTS, A2BS and NNR models all run local on your device!! [MNN TaoAvatar](./apps/Android/Mnn3dAvatar/README.md)
- [2025/06/11] New App MNN TaoAvatar released, you can talk with 3DAvatar offline with LLM, ASR, TTS, A2BS and NNR models all run local on your device!! [MNN TaoAvatar](./apps/Android/Mnn3dAvatar/README.md)
<p align="center">
<img width="20%" alt="Icon" src="https://meta.alicdn.com/data/mnn/avatar/avatar_demo.gif" style="margin: 0 10px;">
</p>
Expand All @@ -24,10 +38,6 @@
<img width="20%" alt="Icon" src="./apps/Android/MnnLlmChat/assets/image_image_new.jpg" style="margin: 0 10px;">
</p>


<details>
<summary> History News </summary>

- [2025/04/30] android app support qwen3 and dark mode [MNN Chat App](./apps/Android/MnnLlmChat/README.md#releases).
<p align="center">
<img width="20%" alt="Icon" src="https://meta.alicdn.com/data/mnn/qwen_3.gif" style="margin: 0 10px;">
Expand Down Expand Up @@ -154,13 +164,13 @@ The group discussions are predominantly Chinese. But we welcome and will help En

Dingtalk discussion groups:

Group #1 (Full): 23329087
Group #4 (Available): 160170007549

Group #2 (Full): 23350225
Group #3 (Full)

Group #3: QR code:
Group #2 (Full): 23350225

![MNN-3](doc/dingdingmnn3.png)
Group #1 (Full): 23329087

## Historical Paper

Expand Down
10 changes: 4 additions & 6 deletions README_CN.md
Original file line number Diff line number Diff line change
Expand Up @@ -111,12 +111,10 @@ MNN适配的硬件架构与精度详见下表:
## 社区交流与反馈
钉钉群组:

- 钉钉群1:23329087
- 钉钉群2:23350225
- 钉钉群3:扫描二维码加入

![MNN-3](doc/dingdingmnn3.png)

- 钉钉群3 (可加入): 160170007549
- 钉钉群3 (已无法加入)
- 钉钉群2 (已满): 23350225
- 钉钉群1 (已满): 23329087

## 历史论文

Expand Down
9 changes: 5 additions & 4 deletions README_JP.md
Original file line number Diff line number Diff line change
Expand Up @@ -117,13 +117,14 @@ MNN(テンソル計算エンジン)に基づいて、推論、トレーニ

Dingtalkディスカッショングループ:

グループ#1(満員):23329087

グループ#2(満員):23350225
グループ#4 :160170007549

グループ#3:QRコード:
グループ#3 (満員)

![MNN-3](doc/dingdingmnn3.png)
グループ#2(満員):23350225

グループ#1(満員):23329087

## 歴史的な論文

Expand Down
Loading
Loading