GithubStar/ncnn
1
0
Fork 0
mirror of https://github.com/Tencent/ncnn.git synced 2026-10-23 17:19:20 +08:00
Code Issues Packages Projects Releases Wiki Activity
3,334 Commits 4 Branches 50 Tags
f1bdc87478c64e0dfdba3d679e6e0dfd4b84df80
Commit Graph

232 Commits

Author SHA1 Message Date
nihui 0734b657d9 spectrogram and inverse spectrogram (#5779) 
* only supports hann, hamming and all-one window
* inverse spectrogram does not support length parameter
* spectrogram always returns torch.view_as_real(out) as ncnn does not support complex typed mat yet
* inverse spectrogram always accepts torch.view_as_complex(in) as ncnn does not support complex typed mat yet
 2024-11-22 16:14:54 +08:00
nihui 66b54cbea2 multiheadattention int8 quantization (#5733) 
* x86 vulkan fallback
* comment about bf16s
 2024-10-15 16:38:00 +08:00
nihui 1c7af00499 gemm int8 quantization (#5706) 
* quantize gemm

* write gemm quantize scales

* update doc

* less openmp args

* x86 riscv fallback

* skip gemm vulkan int8

* fix noint8 test, fix arm bf16 test

* enable vfpv4 on neon build only

* fix gemm vulkan without C

* fp16 pack8 output

* enable elempack=8 only for asimdhp+

* tiled gemm int8 test

* opt arm64 tiles, fix asimdhp dispatch
 2024-10-12 19:02:55 +08:00
Ankush Goel 9b5f6a39b4 fix: typo (#5709) 2024-09-25 08:34:07 +08:00
nihui 5df5413c81 embed int8 quantization and add embed test (#5667) 2024-09-02 18:48:01 +08:00
nihui fdf0df3079 RMSNorm (#5630) 2024-08-15 16:52:06 +08:00
張小凡 051b04ffb4 Updated use-ncnn-with-pytorch-or-onnx document (#5557) 2024-07-24 10:40:17 +08:00
luxincn 02327ba96f add esp32 build document and ci Refs #5536 (#5567) 2024-07-09 11:03:05 +08:00
TianZer fc6b753d31 Add mingw ci and building document (#5547) 2024-07-01 17:55:24 +08:00
nihui 4c3debae2d multiheadattention scale param (#5526) 
* update swiftshader

* skip vs2017 swiftshader
 2024-06-23 23:51:02 +08:00
村长大人 1e75a2df21 add harmonyos how to build with vulkan (#5475) 2024-05-29 21:35:29 +08:00
RoachZhao d4292e9a65 Update vulkan-notes.md (#5472) 
`compute_queue_count` is a function.
 2024-05-25 16:50:39 +08:00
nihui 08b7d99a75 rnn/lstm/gru dynamic quantization (#5435) 2024-05-08 19:25:51 +08:00
Tabbleman be15dbe421 add riscv-gnu-toolchain build guide;-) (#5446) 2024-05-06 22:29:16 +08:00
nihui db035d602d update ncnnoptimize layers, lightmode=false keeps original weight (#5414) 2024-04-09 23:53:54 +08:00
lll143653 342faf2e79 Update how-to-build.md (#5389) 
modify the "bulid for macOS"
 2024-03-27 17:35:09 +08:00
Galasnow 964ed7a56a Add implementation of build for protobuf>=22.0 on Windows (#5359) 
And fix a missing word.
 2024-03-04 11:29:28 +08:00
luqiang-guo 8ddc85f4dd fix doc dual issue (#5342) 2024-02-22 19:13:31 +08:00
afredooo 96d073d541 Some typo fixes (#5339) 2024-02-17 21:12:18 +08:00
hugo-syn 7d8019d577 chore: add markdown code highlight (#5302) 
Signed-off-by: hugo-syn <hugo.vincent@synacktiv.com>
 2024-01-22 12:30:58 +08:00
hugo-syn f35eb4b3b8 chore: Fix multiple typos (#5301) 
Signed-off-by: hugo-syn <hugo.vincent@synacktiv.com>
 2024-01-20 21:54:27 +08:00
nihui 5329d32e74 check vulkan fp16 uniform support and implement lfp conversion without fp16u (#5287) 2024-01-17 15:53:34 +08:00
nihui c222208cc9 feat mask for disable threading, make some extractor setter no-op, update doc (#5270) 2024-01-08 17:42:08 +08:00
Ikko Eltociear Ashimine 5581d27d4d docs: update FAQ-ncnn-vulkan.md (#5268) 
plase -> please
 2024-01-08 10:44:19 +08:00
JeremyRand ed22eb44cc Document libomp-dev dependency (#5228) 
Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>
 2023-12-18 11:24:24 +08:00
lll143653 d4dcb3a2f0 Update FAQ-ncnn-produce-wrong-result.md (#5220) 2023-12-13 17:32:08 +08:00
JeremyRand c1d952da7e Fix "Rasberry" typo (#5182) 
Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>
 2023-11-28 16:04:57 +08:00
JeremyRand 765ac7aef6 Update Vulkan dependency docs (#5178) 
Vulkan deps are optional. vulkan-utils is replaced with vulkan-tools
since Debian 10.

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>
 2023-11-28 10:25:37 +08:00
JeremyRand d1f6193250 Update POWER Clang version docs (#5174) 
Clang prior to 13 no longer fails to build ncnn since #4845.

Clang 18 fixes SSE4.1 translation, which yields a major speedup.

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>
 2023-11-27 10:32:52 +08:00
張小凡 2ecaf37a3e Fix find GPU driver dll path in windows (#5141) 2023-11-10 19:07:24 +08:00
nihui b4f26237cb in-house vulkan loader (#5130) 
* vulkan-driver-loader.md
* static vulkan on apple
 2023-11-09 16:58:23 +08:00
nihui 4494aadd74 deconvolution dynamic weight (#5119) 2023-10-31 17:37:24 +08:00
Justin Fung 009f5eae97 Add description of build for Nintendo 3DS homebrew launcher (#5116) 2023-10-28 21:47:01 +08:00
nihui 14e14a9ae8 slice with indices (#5103) 2023-10-25 17:12:03 +08:00
Yoh 3f437d3f3d Grid sample op (#4373) 
* pnnx support grid_sample op

* complete the permute and gridsample operator fusion

* spilt calculation into two stages and support permute fusion
 2023-10-20 19:09:59 +08:00
nihui 26a70c9b05 fix build with vanilla c906 toolchain (#5048) 2023-09-22 10:47:34 +08:00
Amir Ramezani 7e5fa3ade3 shrink operator (#5022) 2023-09-20 14:08:31 +08:00
Beq Jal bcfec1da33 Celu layer and export to ncnn (#5019) 2023-09-12 10:28:22 +08:00
Beq Jal c851231832 add diag layer and its converter (#4935) 2023-09-08 22:01:37 +08:00
nihui cd1c0d6eab fix build with new protobuf target (#4955) 2023-08-22 14:07:13 +08:00
nihui 4abadd2ffb binaryop implicit broadcast B with 1 dimension rank for outer axis (#4930) 2023-08-10 21:29:49 +08:00
JeremyRand 0a8cf31a05 Add POWER8 VSX toolchains (#4853) 
* Add POWER8 VSX toolchains

POWER8, though slower than POWER9, is still used in the wild; these
toolchains should still be much faster on POWER8 than POWER8 without VSX
optimizations.

* VSX toolchains: set -cpu arg in QEMU CI tests
 2023-08-06 22:16:34 +08:00
mizu-bai 4c861a0d1a Add Building with Intel oneAPI (#4920) 2023-08-06 21:41:12 +08:00
ฅ'ω'ฅ 2303b77ac1 Update how-to-build.md (#4872) 2023-07-21 19:35:36 +08:00
JeremyRand 47e0daf4a1 Translate x86_64 SSE to ppc64le VSX intrinsics (#4807) 
* Add POWER9 VSX toolchains

Translating x86_64 SSE to ppc64le VSX intrinsics yields a quite large
speedup on POWER9. See this article for background:

https://www.talospace.com/2019/07/easier-power-vectorizing-for-fun-and.html

* Add power9le docs

* power9le clang toolchain: Document Clang 13+ requirement

---------

Co-authored-by: Jeremy Rand <jeremyrand@danwin1210.de>
 2023-07-06 16:02:05 +08:00
Kin Yu Shek e8d8042b90 Fix a mistake in docs/faq (#4837) 2023-07-04 08:17:10 +08:00
張小凡 1e0d70af8c Add translated document: glsl-extension.zh.md (#4818) 2023-07-02 11:47:44 +08:00
nihui 43aba6badb Update glsl-extension.md 2023-06-26 11:48:14 +08:00
nihui 172b748c74 add ncnn glsl extension doc (#4817) 2023-06-26 11:13:41 +08:00
nihui 9022b7162a implement all explicit binaryop broadcast types (#4809) 
* simplify binaryop

* less gpu test

* update binaryop broadcast doc

* do not test atan2 zero
 2023-06-19 19:32:32 +08:00