github-featured/alibaba--mnn
2
0
Fork 0
Code Issues Pull Requests Deployments Models Agent Notes Packages Projects Releases Wiki Activity
3,002 Commits 107 Branches 115 Tags
master
Commit Graph

42 Commits

Author SHA1 Message Date
王召德 580ac4da8b [Test:Speed]Add GEMM performance benchmarks for FP32, INT8, and Block32–INT4 (#4516) 
GitOrigin-RevId: c5fad6c80af0c4c29c50b658d8784221dcc654a4
 2026-06-05 17:11:12 +08:00
王召德 8d9f6131ac [CPU:Feature] Support low-bit ARM quant kernels (#4487) 
- Add 2-bit and 3-bit ARMV86 SMMLA kernels for FP16/FP32 paths
- Add 2-bit and 3-bit ARMV82 SDOT kernels for FP16/FP32 paths
- Improve 4-bit and low-bit decode performance paths
- Add LinearRoofline benchmark coverage
- Update ARM CPU optimization skill documentation
- Document MNN_CUDA_NATIVE_ARCH cmake option
- Trim weightBits in lowMemory tests to fix CI timeout

Discussed-in: Merge-Request 27549430 , URL: https://code.alibaba-inc.com/AliNN/AliNNPrivate/codereview/27549430
GitOrigin-RevId: a42c74d902cfdeff92a3b23eb3b4324395944ba7
 2026-05-28 21:04:34 +08:00
Artyom Katolikov 39be21c56f [Infra:Feature] Add test_ci.sh and fix Vulkan/OpenCL op bugs (#4422) 
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
 2026-05-25 20:46:04 +08:00
MNNSyncBot a68e1e3e68 [LLM:Perf] skip fp32 round-trip for fp16 alpha in lowMemory conv init 
Discussed-in: Merge-Request 27442194 , URL: https://code.alibaba-inc.com/AliNN/AliNNPrivate/codereview/27442194
GitOrigin-RevId: 5d45dac0dc14488a2b53e81ae0811caf02d65606
 2026-05-18 15:14:44 +08:00
MNNSyncBot 23d05cc24f [LLM:Feature] support fp16 storage for weight-quant scale/zero-point 
Discussed-in: Merge-Request 27334066 , URL: https://code.alibaba-inc.com/AliNN/AliNNPrivate/codereview/27334066
GitOrigin-RevId: abf0cc65cd38ea1dc346142eb6c91a9775c4e207
 2026-05-18 14:21:37 +08:00
MNNSyncBot 2e0ac31978 [LLM:Feature] add mmap_size in config. 
GitOrigin-RevId: 31bfc11cecb7d56a47b3084b5be9a83f39754e3c
 2026-03-31 17:59:30 +08:00
xiaying 7fde7c7079 MNN:Sync: Sync Internal 3.2.4 2025-09-22 23:05:26 +08:00
zhaode.wzd 55a59a7ebc [MNN:Sync] Sync Internal reranker, gpt-oss. 2025-08-08 12:24:23 +08:00
zhaode.wzd a019d971ad [MNN:Sync] Sync Internal 3.1.4. 2025-05-08 12:39:44 +08:00
xiaying 0769b81b58 MNN:Sync: Sync Internal 3.1.3 2025-04-28 11:50:24 +08:00
xiaying c0247c6998 MNN:Sync: Sync Internal 3.1.1 2025-03-12 11:35:16 +08:00
xiaying 3b6ddc0341 MNN:Sync: Sync Internal 3.0.5 2025-02-12 11:14:19 +08:00
xiaying 27da5e7d48 MNN:Sync: sync internal 3.0.3 2024-12-31 15:34:41 +08:00
xiaying 5b901d9d87 MNN:Sync: Sync Internal 3.0.0 2024-11-18 14:40:27 +08:00
xiaying ae6253fb46 MNN:Sync: Sync Internal 2.9.4 2024-08-24 15:46:21 +08:00
xiaying c6f25cafc6 MNN:Sync: Sync Internal 2.9.3 2024-07-22 20:51:06 +08:00
xiaying 65ec0ea406 MNN:Sync: Fix bug for llama2/llama3 attention fuse, refract llm usage 2024-06-15 15:39:59 +08:00
xiaying 7cad2ee83f MNN:Sync Sync Internal 2.9.0 2024-05-11 19:17:02 +08:00
xiaying 5895243607 [MNN:Sync] Sync Internal 2.8.4 2024-04-19 11:58:21 +08:00
zhaode.wzd 2972fe71dc [MNN:Sync] Sync Internal 2.8.3 2024-03-13 14:55:54 +08:00
zhaode.wzd 970b63f3b4 [MNN:Sync] Sync Internal 2.8.2 2024-02-29 16:21:40 +08:00
zhaode.wzd bdf15442f4 [MNN:Sync] Sync Internal 2.7.1 2023-09-21 09:01:35 +08:00
xiaying 98ba00c2f3 [MNN:Sync] Sync Internal 2.6.3 2023-08-21 14:53:27 +08:00
zhaode.wzd 67eceb8abb [MNN:Sync] Sync Internal code, support low_memory for conv. 2023-06-27 10:33:16 +08:00
xiaying c70ecef660 [MNN:Sync] Sync Internal Gitlab: 2.5.1 2023-05-18 19:11:50 +08:00
xiaying 4a609006eb [MNN:Sync] Sync Internal 2.3.1 2023-02-15 10:30:27 +08:00
xiaying d46b6b998d [MNN:Sync] Sync Internal 2.3.0 2022-12-30 15:18:58 +08:00
xiaying 68708c5d66 Sync Internal 2.0.4 2022-08-12 10:30:48 +08:00
xiaying 3a0fb480d2 Fix crash bug for origin quan model 2021-09-06 17:18:29 +08:00
tianhang.yth 4eb1096b9c fix alpha div by zero bug and arm server compile bug 2021-06-24 10:38:55 +08:00
hush-alibaba 58545d6ca1 Synchronize internal github for version 1.2.0 (#1518) 2021-06-11 17:17:13 +08:00
tianhang.yth d85952d826 sync from internal repo 2021-04-28 18:02:10 +08:00
雁行 453264cb5e [PATCH 6/7] [QAUNT:Bugfix] Bugfix for IDST encode when weight value = 1. 2021-04-22 13:51:14 +08:00
弗人 9621a1f8ad [PATCH 14/36] bug fix for old type 4 models 2021-04-16 14:29:36 +08:00
弗人 a3bbcd01b9 [PATCH 12/36] [Train:Featue] support full quant for train quant, encode when save 2021-04-16 14:29:36 +08:00
弗人 95bcb842a0 [PATCH 08/36] [Train:Feature:Bugfix] train quant support full quant 2021-04-16 14:29:36 +08:00
xiaying d91fc63976 [MNN:Sync] Sync internal Gitlab 2021-04-08 15:34:23 +08:00
xiaying 2d1b129121 [MNN:Sync] Sync internal git 2021-01-06 16:29:37 +08:00
xiaying 29a39f1c10 [MNN:Sync] Sync internal gitlab 2020-11-25 19:03:07 +08:00
Hui Shu d6795ad031 Github release 1.1.0 2020-11-05 16:49:17 +08:00
xiaying dbd3267c09 [PATCH 114/160] [MNN:Refractor] Seperate the pad compute for convolution and deconvolution 2020-03-22 19:02:11 +08:00
xiaying 010913aaaa [PATCH 070/160] [MNN:Refract] Seperate the load of quan and half to ConvolutionCommon 2020-03-22 19:02:06 +08:00