8b7d8d29
fix(core): fix json dump when weight preprocess by
2021-05-24 19:39:02 +0800
ec65e1f9
fix(build/windows): fix windows build: * compat clang-cl 11 build at windows env * fix cuda/cudnn/trt copy env build failed on windows by
2021-05-25 23:00:55 +0800
8a918717
feat(mgb): add megengine inference by
2021-05-25 15:21:24 +0800
379a28f9
feat(mgb): add load_and_run option by
2021-05-21 17:21:17 +0800
b06b5899
feat(mgb): get static graph memory info by
2021-04-23 17:20:53 +0800
0cf4ff70
perf(dist): add fastpath for bcast params by
2021-05-26 13:37:05 +0800
5e548218
docs(syncbn): complete syncbn document by
2021-05-24 15:01:27 +0800
3591ef1f
fix(mgb): fix conv cudnnconvbackwarddata algo witch is not shake by
2021-05-21 21:59:04 +0800
e417798f
fix(mge): correct pytype when calling apply from python by
2021-05-21 18:24:25 +0800
c4048519
chore(mge/version): change dev version to 1.5 by
2021-05-24 10:52:35 +0800
3352fe70
refactor(dtr): import dtr as submodule by
2021-05-21 10:42:37 +0800
a7581986
perf(dist): speed up bcast_val by
2021-05-21 02:45:23 +0800
1525a025
feat(mge/module): add python wrapper for unfold by
2021-05-10 17:14:32 +0800
d07ef83a
feat(mge/tools): add bash_completion for user to hint script by
2021-05-19 16:10:14 +0800
efc7f04f
refactor(mge/jit): using static global_enable for apply ctx insted of global variable by
2021-05-13 12:53:01 +0800
750bde95
refactor(mge/jit): remove is_compiled flag in cpp tensor by
2021-05-13 13:41:59 +0800
4f240ec2
refactor(mge/jit): make trace return any kind of output by
2021-04-06 15:09:40 +0800
c6b552cf
refactor(mge/functional): remove dependence to trace in functional implementations by
2021-05-12 17:20:49 +0800
46d96478
feat(mge/tools): optimize statistical tools by
2021-05-19 19:59:50 +0800
7f9d441c
docs(mge/module): add note about weight shape in conv by
2021-05-13 15:50:52 +0800
8be70d17
docs(misc): correct docstring format broadly by
2021-05-12 08:39:08 +0800
53446bce
fix(imperative/dataset): fix VOC dataset by
2021-05-14 14:45:11 +0800
8e5bf948
fix(mge/utils): fix bug of VarNode inplace operations by
2021-05-12 19:36:52 +0800
13b15fb0
feat(megbrain): add correlation opr by
2021-05-13 14:02:16 +0800
1997b1a2
feat(dnn/cuda): add correlation kernel by
2021-04-26 13:55:27 +0800
6d376623
fix(imperative/dataloader): collect garbage explicitly before starting a process in dataloader by
2021-05-19 15:57:34 +0800
25989425
feat(imperative/autodiff): add grad clip by
2021-04-15 18:57:56 +0800
601a33a8
refactor(mge/dtr): update dtr api by
2021-05-07 11:34:39 +0800
c269a690
fix(mge/dtr): filter bad ops by
2021-05-18 11:33:18 +0800
b230e146
fix(mge/dump): fix dump_with_testcase_mge Varnode type mismatch by
2021-05-11 19:37:25 +0800
9b69a02f
fix(imperative/tensor): init m_offset when constructing a Tensor with DeviceTensorND by
2021-03-16 11:59:37 +0800
acb000d0
fix(api_cache): fix serialization for conv_desc by
2021-05-17 19:32:31 +0800
c4032222
fix(api_cache): lock api cache for thread safety by
2021-05-12 14:35:56 +0800
5419a95d
perf(cuda/conv): cache serval cudnn api by
2021-03-16 19:37:30 +0800
19887942
feat(dnn/apicache): add generic apicache by
2021-03-16 19:32:06 +0800
226f24fa
test(mge/distributed): add gather scatter reduce broadcast grad test by
2021-04-15 19:44:58 +0800
aea5ac13
fix(mge/distributed): fix gather scatter reduce broadcast autodiff by
2021-04-15 19:28:02 +0800
b1bf193e
feat(functional/loss): add reduction choices to loss functions by
2021-01-27 14:16:51 +0800
36b1ba05
fix(mgb/dnn): fix cudnn8.0.4 convbias with z by
2021-03-08 20:58:54 +0800
dadd5086
test(param_pack): more cases for param pack concat by
2021-04-28 16:52:13 +0800
346d2420
fix(param_pack): impl param pack concat in imperative_rt by
2021-04-28 16:00:07 +0800
579a3d77
fix(mge/jit): error out if dump bn in training mode by
2021-05-17 12:44:50 +0800
68da973e
fix(mge/network_visualize): fix warning 'span dist too large' by
2021-05-14 17:40:18 +0800
baa30400
feat(build): change make to ninja by
2021-05-11 06:35:25 +0800
47138c06
(tag: v1.4.0, release-1.4)
perf(dist): add fastpath for bcast params by
2021-05-26 13:37:05 +0800
1a21dfdf
refactor(dtr): import dtr as submodule by
2021-05-21 10:42:37 +0800
6964576e
perf(dist): speed up bcast_val by
2021-05-21 02:45:23 +0800
052a600f
feat(mge/module): add python wrapper for unfold by
2021-05-10 17:14:32 +0800
77a4bebf
feat(mge/tools): add bash_completion for user to hint script by
2021-05-19 16:10:14 +0800
bff7d163
refactor(mge/jit): using static global_enable for apply ctx insted of global variable by
2021-05-13 12:53:01 +0800
2b7da422
refactor(mge/jit): remove is_compiled flag in cpp tensor by
2021-05-13 13:41:59 +0800
2762a552
refactor(mge/jit): make trace return any kind of output by
2021-04-06 15:09:40 +0800
a29bf679
refactor(mge/functional): remove dependence to trace in functional implementations by
2021-05-12 17:20:49 +0800
2621c378
feat(mge/tools): optimize statistical tools by
2021-05-19 19:59:50 +0800
fb69697f
docs(mge/module): add note about weight shape in conv by
2021-05-13 15:50:52 +0800
186768e3
docs(misc): correct docstring format broadly by
2021-05-12 08:39:08 +0800
aa0e11fe
fix(imperative/dataset): fix VOC dataset by
2021-05-14 14:45:11 +0800
2c2beac9
fix(mge/utils): fix bug of VarNode inplace operations by
2021-05-12 19:36:52 +0800
4fa9096d
feat(megbrain): add correlation opr by
2021-05-13 14:02:16 +0800
b3aae4bb
feat(dnn/cuda): add correlation kernel by
2021-04-26 13:55:27 +0800
866ab6e6
fix(imperative/dataloader): collect garbage explicitly before starting a process in dataloader by
2021-05-19 15:57:34 +0800
67167cb3
feat(imperative/autodiff): add grad clip by
2021-04-15 18:57:56 +0800
d4dd134d
refactor(mge/dtr): update dtr api by
2021-05-07 11:34:39 +0800
9122ff2e
fix(mge/dtr): filter bad ops by
2021-05-18 11:33:18 +0800
71e007c3
fix(mge/dump): fix dump_with_testcase_mge Varnode type mismatch by
2021-05-11 19:37:25 +0800
7d181027
fix(imperative/tensor): init m_offset when constructing a Tensor with DeviceTensorND by
2021-03-16 11:59:37 +0800
af62add0
fix(api_cache): fix serialization for conv_desc by
2021-05-17 19:32:31 +0800
d0a68c43
fix(api_cache): lock api cache for thread safety by
2021-05-12 14:35:56 +0800
e8061e66
perf(cuda/conv): cache serval cudnn api by
2021-03-16 19:37:30 +0800
fe8488bf
feat(dnn/apicache): add generic apicache by
2021-03-16 19:32:06 +0800
45b9b5ed
test(mge/distributed): add gather scatter reduce broadcast grad test by
2021-04-15 19:44:58 +0800
78d1e2b2
fix(mge/distributed): fix gather scatter reduce broadcast autodiff by
2021-04-15 19:28:02 +0800
27ff38dd
feat(functional/loss): add reduction choices to loss functions by
2021-01-27 14:16:51 +0800
927369da
fix(mgb/dnn): fix cudnn8.0.4 convbias with z by
2021-03-08 20:58:54 +0800
0a1bd90f
test(param_pack): more cases for param pack concat by
2021-04-28 16:52:13 +0800
49a8134d
fix(param_pack): impl param pack concat in imperative_rt by
2021-04-28 16:00:07 +0800
acb23992
fix(mge/jit): error out if dump bn in training mode by
2021-05-17 12:44:50 +0800
dc12d319
fix(mge/network_visualize): fix warning 'span dist too large' by
2021-05-14 17:40:18 +0800
2f1e7c53
feat(build): change make to ninja by
2021-05-11 06:35:25 +0800
e70c9a38
fix(cmake): fix cmake depends reason: some target call python3 to generate some headers, if PYTHON_EXECUTABLE changed, headers will be invalid, which will lead to rebuild most of obj, because this headers will be include in most cpp/c, what`s more, this target do not depends python3 version(python API) the output is same when use python3.5 or python3.x so change use PYTHON_EXECUTABLE to PYTHON3_EXECUTABLE_WITHOUT_VERSION, when PYTHON_EXECUTABLE/PYTHON_LIBRARY/PYTHON_INCLUDE_DIR changed, can reuse the build obj by
2021-05-10 13:50:56 +0800
e5d8e427
feat(mge): add a tool which can analyze the file generated by compare_binary_iodump.py by
2021-05-10 15:42:25 +0800
0678a169
feat(mgb): load_and_run can set both fastrun and reproducible by
2021-05-10 15:37:22 +0800
747b53c0
feat(dnn): add conv_bwd_data and conv_bwd_filter accuracy shake check by
2021-04-27 16:55:26 +0800
a405461c
fix(mge/elemwise): fix problem that elemwise.mode is not comparable with string mode by
2021-05-12 11:51:19 +0800
1aba1918
fix(imperative/ops): add check_dtype for Elemwise in infer_attrs by
2021-05-11 17:36:06 +0800
53539fed
fix(mge/quantization): fix a compute problem by
2021-05-11 16:58:09 +0800
94b5e1b0
fix(mge/core): support non-Tensor args for `core.tensor.utils.get_device` by
2021-05-11 14:23:28 +0800
609b37b9
perf(fastrun): cache persistent cache by
2021-05-12 17:05:04 +0800
0368eff6
fix(mgb): fix fastrun workspace limit by
2021-05-12 14:26:50 +0800
b6dc4c82
fix(tensor): del valid tensors when compnode finalizing by
2021-03-18 13:17:39 +0800
1ecf2ab4
fix(mge/autodiff): check tensors to be attached by
2021-05-13 14:56:53 +0800
2c61b8bf
feat(mge/functional): argmin and argmax support negtive axis by
2021-05-05 11:44:49 +0800
cb84efa7
fix(mge): fix dumping backward graph by
2021-05-06 15:37:26 +0800
bb94f2aa
feat(mge/module): add __repr__ method for qat and quantized module by
2021-03-30 14:45:06 +0800
5dbf3612
test(mge/optimizer): update optimizer test to make sure grad not change by
2021-04-23 17:11:13 +0800
e9d6361e
fix(mge/optimizer): fix optimizer update step inplace add grad by
2021-04-23 16:24:56 +0800
a3e098c8
fix(mge/utils): fix network multiple outputs issue by
2021-05-06 18:26:31 +0800
ecb11202
fix(build): fix naive build by
2021-04-29 17:24:33 +0800
ce7913e3
fix(mge/imperative): fix proxy graph comp_node finalize by
2020-10-14 19:11:08 +0800
0d332cf0
feat(imperative/opr): rebase rng refactoring to dev & add python module by
2021-02-24 21:16:33 +0800