042c7fd1
(try-import)
feat(dnn): add elemwise modes by
2022-06-14 12:38:17 +0800
9889e82e
(HEAD -> master)
docs(imperative): add example of dataset by
2022-11-14 14:18:21 +0800
7d83a9ad
fix(imperative): infinite sampler support get batchsize by
2022-11-14 11:51:26 +0800
fa488338
feat(dnn,imperative): remove the restriction of tensor shape when using uint8 region mask by
2022-10-26 17:30:02 +0800
cdf7baa2
feat(python): support python 3.10 by
2022-10-17 19:11:17 +0800
30ef28b0
build(imperative): generate stub files for core ops by
2022-11-04 06:29:46 +0000
8fe5e74f
Revert "fix(imperative): fix the matrix mul error when symbolic trace" by
2022-11-15 12:38:30 +0800
7f280058
fix(imperative): fix the matrix mul error when symbolic trace by
2022-11-09 19:45:53 +0800
485f0a1f
fix(imperative): fix redis fastrun cache server by
2022-11-10 16:59:20 +0800
551cad49
refactor(megbrain): refactor try infer tensor layout in lite avoiding using megbrain interface by
2022-07-05 14:35:35 +0800
18d7a97c
fix(jit/cpu): make lar -enable_jit args possible on other backends, for example cpu with MLIR by
2022-11-07 15:05:20 +0800
63406bad
fix(ycm): make ycm happy with llvm by
2022-11-04 17:15:31 +0800
68d7320e
fix(mge/utils): fix using wrong function in register_hook_module by
2022-11-02 18:24:45 +0800
3018ca51
fix(mge): add param RuntimeArgs to customop kernel on cuda by
2022-06-28 15:06:12 +0800
086ee045
fix(mge): fix error when transform model to nchw4/32/64 tensor format by
2022-10-18 17:04:00 +0800
841bfe92
fix(build): make develep happy by
2022-11-08 17:05:44 +0800
38ef331e
Merge pull request #492 from tpoisonooo:fix-imagenet-untar by
2022-11-08 19:33:35 +0800
2ed38f45
feat(test): add some comment and test in imperative by
2022-11-01 18:59:38 +0800
0ebd4400
fix(dnn): fix the modulo of int by
2022-11-03 14:21:40 +0800
6fac92dd
feat(whl): force import megbrain before import megengine by
2022-11-01 21:06:06 +0800
68d27108
fix(build): remove build so many warning on windows by
2022-11-03 18:37:40 +0800
4207dc7b
feat(whl): check broken env before import 2/2 by
2022-11-01 17:40:56 +0800
727be7ea
docs(imperative): adjust bn doc by
2022-11-01 15:31:05 +0800
1137ebf3
fix(imperative): fix the tensor info lock in worker by
2022-10-31 10:13:06 +0800
2694ff81
fix(mge/module): fix some deconv fuse bn problem by
2022-10-25 19:11:11 +0800
1886ebc3
fix(lite/load_and_run): fix fast-run invalid in fitting mode by
2022-10-27 18:53:02 +0800
20719502
fix(imperative): fix matmul deduce dtype by
2022-11-02 13:59:25 +0800
d733f429
feat(whl): check broken env before import by
2022-10-31 19:09:59 +0800
b4532af8
fix(lite/load_and_run): fix log info for ads profile_support by
2022-10-25 11:09:01 +0800
582dd4ce
fix(dnn/sfotmax): call cpu dispatch for softmax opr by
2022-10-28 16:53:36 +0800
2976f60b
fix(test): tmp skip groupnorm test by
2022-10-28 18:56:16 +0800
235d81dd
feat(dnn): add fp16 nchw88 im2col algo by
2022-10-09 10:32:25 +0800
f7d2017e
fix(serialization): add tensor value loader support in new format by
2022-10-25 18:39:50 +0800
da7f250c
feat(mge/module): deconv fuse bn and relu by
2022-08-18 15:34:16 +0800
dbd94839
feat(dnn,src,imperative): add groupnorm op by
2022-08-30 14:28:43 +0800
cc6084b4
(tag: v1.11.1, release-1.11.1)
revert: chore(mge/misc): api converage by
2022-10-13 19:18:56 +0800
069e4e07
fix(mge/utils): fix module_stats missing param problem by
2022-10-19 20:20:52 +0800
a0d36566
fix(mge/functional): fix memory leak when set tensor dtype to Quantized by
2022-09-26 13:51:22 +0800
f01b1255
fix(mge/functional): fix illegal memory access by
2022-09-26 13:50:27 +0800
95b2a36f
feat(core): add zero copy input output from out of megengine by
2022-10-09 15:45:13 +0800
563a0d01
feat(opencl/ion): support ion by
2022-09-29 19:09:04 +0800
6d219827
feat(opencl/svm): add opencl svm more test by
2022-09-05 21:58:38 +0800
4c27522a
fix(src/core): fix weight preprocess free combine memory UAF issue by
2022-09-05 12:51:08 +0800
0a52b258
fix(opencl/test): fix test weight preprocess filter UAF issue by
2022-09-02 11:02:00 +0800
8fe8edf4
feat(dnn): add fp16 mk8 16x12 matmul algo by
2022-10-09 10:32:25 +0800
21a975f8
feat(gopt): add channel padding pass when enable nchw44/nchw88/nchw44-dot by
2022-09-20 17:19:12 +0800
b82e8f00
refactor(gopt): refact the padding channel opt pass by
2022-09-16 18:41:11 +0800
f444d4fe
feat(dnn,imperative): region restricted conv support groups=1 even if depthwise by
2022-10-17 16:31:13 +0800
d14ae856
fix(imperative): update region restricted conv testcase for bias shape by
2022-10-17 15:05:37 +0800
ffb62624
ci(imperative): fix imperative test_ctc_loss test by
2022-10-17 19:08:36 +0800
2a4295fb
feat(ci): update submodules by
2022-10-09 11:43:47 +0800
75c413e3
feat(imperative): region restrictd conv support bias in python by
2022-10-09 16:44:51 +0800
6f9f25a8
fix(gopt): fix global layout transform fold conv typecvt by
2022-10-15 12:32:52 +0800
fa9d719f
fix(gopt): fix global layout transform fold conv typecvt by
2022-10-15 12:32:52 +0800
cc9f743e
fix(serilization): fix elemwise multitype compatibilty in v1.11.0 by
2022-10-14 16:54:21 +0800
767fa474
fix(mgb): add detail error message for trt version diff by
2022-09-30 14:05:00 +0800
94e49fae
fix(serilization): fix elemwise multitype compatibilty in v1.11.0 by
2022-10-14 16:54:21 +0800
0aeabd6b
Merge pull request #488 from is-shidian:python_whl_arm by
2022-10-15 20:33:33 +0800
8f2390fc
fix(imperative): untar imagenet use bad path by
2022-10-15 11:41:09 +0800
8cf1de4a
fix pyenv python version,pyenv install,file location,file name,and compat for x86_64 and arm64 by
2022-09-25 08:29:33 +0800
34c529b3
fix(subgraph): disable jit due to environment issue by
2022-10-12 12:29:00 +0800
194201a7
revert: chore(mge/misc): api converage by
2022-10-13 19:18:56 +0800
dfa999ee
fix(doc): fix warp_affine docstring by
2022-10-13 14:56:22 +0800
ece454fd
fix(third_party): fix cpuinfo related to sve2 by
2022-10-10 19:19:04 +0800
c1c1d6d1
feat(imperative): region restrictd conv support bias in python by
2022-10-09 16:44:51 +0800
f287501e
fix(src/core): fix pushdown alloc_result inconsistent when run for many time by
2022-08-22 20:52:28 +0800
edea4b80
ci(docs): add document build ci logic by
2022-09-21 15:27:45 +0800
5c84305f
fix(imperative): fix broadcast extra assert of output shape by
2022-10-09 16:01:11 +0800
7790040e
fix(ci): fix integration_test bugs because of dataloader by
2022-10-09 10:59:58 +0800
8d4bc094
feat(ci): update submodules by
2022-10-09 11:43:47 +0800
7b3f9e85
chore(release): bump version by
2022-10-08 12:58:50 +0800
6db4620e
feat(dnn): fix wgrad rrconv for compute capability by
2022-10-02 20:30:58 +0800
4e9b1c4e
feat(dnn): add rrconv wgrad, support int32 and uint8 region mask by
2022-09-21 10:27:39 +0800
9e020d23
feat(imperative,mgb): add RegionRestrictedConvolution megbrain and imperative opr by
2022-07-25 13:58:39 +0800
977c2071
feat(dnn): add RegionRestrictedConv DGRAD support int32 and uint8 by
2022-08-09 10:12:38 +0800
543c9b77
feat(dnn): add RegionRestrictedConv cuda by
2022-07-04 19:14:34 +0800
fdec82ec
feat(dnn): add naive RegionRestrictedConv by
2022-07-01 20:35:30 +0800
edc92ccf
perf(imperative/data): improve dataloader preformance by
2022-08-22 20:26:46 +0800
896b0193
fix(mgb): fix zero copy error when model end with memory forward opr by
2022-09-26 19:08:13 +0800
babecba2
fix(mge/traced_module): fix visualization bug by adding Node to non-const leaf by
2022-09-28 17:41:35 +0800
034c7787
fix(mge/functional): support non-float32 input when call isxxx by
2022-09-27 20:18:55 +0800
e9cc5237
fix(mgb): format code by
2022-09-27 10:19:57 +0800
a07fbf79
Merge pull request #484 from wangxiang9603:add-nchw44-deconv by
2022-10-08 12:04:25 +0800
d3b2b519
feat(mge/opr): add meshgrid opr by
2022-09-09 16:32:07 +0800
ec234135
feat(lite): support discrete inputs by
2022-08-26 19:43:50 +0800
dc0ab9b6
feat(lite): replace warp when src is discrete input by
2022-08-08 15:25:07 +0800
58b682ca
feat(dnn/cuda): add naive bmm by
2022-04-25 20:44:27 +0800
eddb0aba
docs(mge/functional): enhance tensor creation function docstring by
2022-09-19 12:32:56 +0800
36d3c2ac
docs(mge/tensor): add dtype and device doc ref for typehint result by
2022-09-19 11:38:49 +0800
22286579
docs(mge/functional): enhance functional statstic func docstring by
2022-09-16 14:41:22 +0800
213f4043
chore(mge/functional): add normalize interace to f.nn namespace by
2022-09-16 16:39:18 +0800
f62f2615
docs(mge/functional): enhance functional elemwise docstring by
2022-09-14 18:10:45 +0800
07d94c41
docs(mge/tensor): add docstring for some tensor attr by
2022-09-14 16:24:43 +0800
554994f0
docs(mge/serilize): enhance megengine save and load docstring by
2022-09-14 15:27:07 +0800
b50947d2
docs(mge/module): add Module.state_dict docstring by
2022-09-14 15:26:23 +0800
edd3ee67
fix(mgb): add error infomation for old version load new elemwise mode by
2022-09-23 19:54:19 +0800
a7e28ebe
fix(dnn): fix winograd load error and cpuinfo test error by
2022-09-19 16:19:06 +0800
41b9db85
fix(mgb): make error infomation of advanced indexing out of bound more readable by
2022-09-16 11:49:37 +0800
eb512b62
fix(mgb/imperative): fix the future_error caused by repeat get calls at the same std::future by
2022-09-22 19:12:44 +0800
2efedf42
fix(mge/tensor): modifiy tensor print suppress options by
2022-09-21 16:29:03 +0800