09b5f3d4
fix(mgb/core): fix multi thread pool deactive and multi thread conflict by
2020-08-04 15:04:54 +0800
ef239f83
feat(windows/python_whl): make windows HAPPY for build megbrain python package by
2020-08-06 15:10:26 +0800
bf6cbc1d
build(third_party): fix git apply issue by
2020-08-24 22:03:09 +0800
5eb491c5
Merge pull request #74 from ChaiMind:master by
2020-08-12 11:34:19 +0800
b72f1e82
chore(build): cleanup BUILD files by
2020-07-16 14:45:32 +0800
e258812f
feat(dnn): add bool dtype by
2020-07-22 19:01:26 +0800
734c498d
perf(mgb/core): improve DevMemAlloc when it has single stream by
2020-07-30 16:05:08 +0800
39bd66fc
fix(mgb): fix TensorRT missing cudaSetDevice by
2020-08-05 21:20:25 +0800
ab9dfbce
test(mgb): fix tensorrt tests missing cudaSetDevice by
2020-08-05 16:24:15 +0800
b43fb1a9
perf(mgb): add CUDA host memory allocator by
2020-07-29 14:48:54 +0800
2afceb41
fix(mgb/atlas): use dyn output alloc if enable dynamic batchsize by
2020-07-30 00:31:29 +0800
6bcc6fae
feat(mge/imperative/opr): modify batch_norm to support frozen BN by
2020-07-23 14:21:54 +0800
7ca3d579
feat(dnn): make mk4 and mk8 matmul for winograd both on aarch64 and armv7 supports n=1 by
2020-07-16 13:27:13 +0800
54d18115
fix(imperative): fix grad of BatchNorm by
2020-08-04 15:12:49 +0800
80c47053
perf(mgb): use midout in megbrain to reduce binary size by
2020-07-21 16:11:58 +0800
35c71276
fix(mge/quant): fix TQT epoch scale change bug by
2020-08-04 16:46:23 +0800
e6e41242
fix(mge/quant): fix zero grad warn in TQT train by
2020-08-03 15:20:12 +0800
4a01d236
ci(mgb/python_module): fix doctest syntax error by
2020-08-05 10:53:08 +0800
b3ac87e4
chore(pip): update pip version by
2020-08-11 23:30:09 +0800
c0528b24
chore(third_party): add cpuinfo by
2020-08-11 13:03:26 +0800
4348960c
fix(mge/gopt): fix fp16 compute mode by
2020-07-29 17:36:27 +0800
9f4060b0
fix(mgb/gopt): add ShuffleShuffleRemovePass assert by
2020-07-31 10:57:03 +0800
4495c0cc
feat(mgb/quantization): add get quantize parameters support by
2020-07-17 15:04:07 +0800
9b097859
chore(core): remove index_sequence polyfill for pre-C++14 by
2020-07-20 18:48:42 +0800
074645d8
feat(core): add array_skip by
2020-07-21 20:11:39 +0800
2faab90b
feat(core): add a naive if_constexpr for C++14 by
2020-07-20 20:35:11 +0800
8f42fe0d
feat(core): import a std::apply polyfill by
2020-07-20 18:46:57 +0800
be5dde9d
fix(mgb/opr-mm): fix group manager sync error by
2020-07-29 13:07:36 +0800
f6018422
perf(dnn/arm_common): add nchw44 winograd f73 by
2020-07-10 16:14:53 +0800
2dc91900
fix(fastrun/persistent_cache): fix fastrun crash by
2020-07-29 19:12:05 +0800
a9e7a670
fix(sdk/c-opr-loaders): fix mace dump model and env params and add tuning by
2020-07-29 11:37:03 +0800
c1fb3c8c
fix(mgb/gopt): fix ConvertF32ToF16Pass endpoints by
2020-07-28 18:03:09 +0800
380cb6e4
feat(mge/jit): add support output symbol var name settings for dump by
2020-07-23 16:44:30 +0800
e1e56988
feat(dnn/fallback): add conv1x1 filter preprocess funciton by
2020-07-08 20:42:24 +0800
0f9dec68
feat(mge/imperative): name so lib by
2020-07-28 18:48:26 +0800
618faf64
feat(mgb/profiler): dump static infer info by
2020-07-15 20:39:09 +0800
e05c795b
refactor(dnn/arm): refactor direct algo in algo selection by
2020-07-20 14:20:30 +0800
134a1026
fix(mge/doc): fix typos, format by
2020-07-27 18:18:06 +0800
324af878
feat(dnn/arm): add cpuinfo runtime check for x86 and arm by
2020-07-26 22:03:22 +0800
8b183f2c
test(dnn/testcase): fix a testcase bug by
2020-07-21 14:37:48 +0800
5c224c71
refactor(mge/imperative): drop ComputingGraphImpl in proxy graph by
2020-07-14 18:52:17 +0800
14a32ae1
fix(cmake/cross-build): misc fix 1: fix cmake cross-ios failed caused by df118a87 build static lib for APPLE define for XCODE third_party framework including require by
2020-07-21 14:19:07 +0800
5e7d2a91
refactor(mgb): add TensorND::proxy_to_default_cpu by
2020-07-16 15:58:59 +0800
7a8a2830
feat(quant): support nnie quant by
2020-06-28 16:33:53 +0800
d49a2971
Update prepare.sh by
2020-08-04 23:48:38 +0800
3f6ae833
build(mge): set right flags for GNU gold linker via MegEngine#44 by
2020-08-04 15:18:52 +0800
02783250
fix(python/import/macos): fix some macos user import crash by
2020-07-24 11:22:39 +0800
5133d9a3
Merge pull request #62 from ChaiMind:patch by
2020-08-04 13:41:54 +0800
edd7e167
feat(dnn/fallback): add im2col filterpreprocess function by
2020-07-02 10:11:07 +0800
9e9e8ca0
fix(mgb): fix some compile errors by
2020-07-15 15:18:09 +0800
ef267dac
fix(megdnn_test/ev300): try run megdnn_test on ev300 board by
2020-07-23 15:54:09 +0800
82692159
fix(mge/module): fix named_children of Sequential by
2020-07-20 14:17:00 +0800
eed54081
feat(dnn/arm): add armv7 mk4 i8i8i16 gemm, optimized for A7 by
2020-07-23 13:28:15 +0800
9c475fff
fix(dnn/fallback): delete ConvBias* opr param of conv_bias algo by
2020-07-10 18:18:18 +0800
bb6030ff
(tag: v0.6.0, release-0.6)
chore(release): bump version by
2020-07-23 16:01:21 +0800
ba5a43b8
fix(dnn/fallback): delete ConvBias* opr param of conv_bias algo by
2020-07-10 18:18:18 +0800
d782edf8
refactor(mgb): decouple node insertion from ComputingGraphImpl by
2020-07-11 18:28:22 +0800
d42cf4cd
refactor(mgb): replace static_cast<ComputingGraphImpl*> with a checked version by
2020-07-11 18:28:11 +0800
4d56371e
refactor(dnn/arm): split arm direct kernel to cut compile time by
2020-07-10 10:24:52 +0800
55844d3e
fix(core): fix CPU compnode get_ram_status_byte free_memory error by
2020-06-29 14:19:07 +0800
672d4ad0
fix(jit): more testcases on the grad of JITExecutor by
2020-06-15 13:58:46 +0800
bc95e873
fix(jit): fix jit grad by
2020-06-11 21:52:07 +0800
fc1ce273
fix(dnn/cuda): fix elemwise add cuda int8 bcast by
2020-07-16 15:21:59 +0800
57bc3657
style(dnn/cuda): format cuda elemwise code by
2020-07-16 15:16:41 +0800
09eaa398
fix(mgb/dnn): fix case fallthrough compile error for gcc7 by
2020-07-15 16:13:24 +0800
b689b4ce
fix(mge/functional): add F.clamp to __all__ to fix docs by
2020-07-20 13:00:48 +0800
4cf193ab
fix(mgb/opr-mm): fix device id in TestOprIORemote.IdentityWithGopt by
2020-07-21 10:23:17 +0800
c76789a5
feat(mge/imperative): add io remote wrapper by
2020-07-15 11:40:22 +0800
d0d7485a
fix(mge/build): update cmake_minimum_required VERSION by
2020-07-14 16:34:46 +0800
9006d76b
fix(serialization): do not gate enum entries by macro by
2020-07-10 17:14:13 +0800
ad4928c9
fix(serialization): remove #include flatbuffers/flatbuffers.h in public header by
2020-07-10 17:03:46 +0800
177c2688
feat(osx/python_whl): fix python pytest failedon osx by
2020-07-14 13:26:40 +0800
d35c609d
feat(osx/python_whl): refine whl doc and macos whl env by
2020-07-14 13:31:27 +0800
4d541b72
feat(osx/python_whl): make macos HAPPY for build megbrain python package by
2020-07-09 18:04:43 +0800
3bfb704e
fix(mgb): fix windows host build type error by
2020-07-17 13:21:30 +0800
ab3efab0
fix(mgb): add compute capability check by
2020-07-15 13:04:25 +0800
8070f40a
fix(mgb/gopt): fix gopt nchwxx convert elemwise and reshape by
2020-07-15 20:12:25 +0800
b38e8225
refactor(mgb/opr-mm): update megray communicator init interface and fix ci by
2020-07-18 21:17:52 +0800
5e912edd
fix(mgb/opr-mm): fix grad func of reduce and gather by
2020-07-18 14:30:04 +0800
e3e981cc
test(mge/collective_comm): fix collective_comm test and add data parallel test by
2020-07-15 17:05:44 +0800
d3d9018f
feat(imperative/opr-mm): add broadcast by
2020-07-14 17:33:04 +0800
b2f05bf8
fix(mge/module): fix quantized fold weight value range limit for fused conv/bn modules by
2020-07-13 15:10:28 +0800
0df74604
fix(data): fix contiguous id by
2020-07-16 17:09:53 +0800
fff2cdc7
feat(dnn/fallback): add winograd weight preprocess by
2020-06-24 19:15:17 +0800
d37229fa
feat(dnn): optimize f23 and f63 nchw44 winograd by
2020-07-03 14:59:24 +0800
d7c0dd45
build(ci/docker_env): add cuda version chek when build megbrain wheel, must using CUDA-10.1 CUDNN-V7.6.3 TENSORRT-6.0.1.5 by
2020-07-01 17:05:50 +0800
3bd8ef35
feat(mgb/compnode): add atlas compnode by
2020-03-31 21:19:17 +0800
aa147b74
fix(osx/python_module): fix build python module on osx by
2020-07-09 14:47:25 +0800
1e576e32
feat(dnn/aarch64-arm_common): add mat_idx warppespective for aarch64/arm_common/naive by
2020-06-22 14:41:30 +0800
714cb232
feat(dnn): add gemv supports in conv1x1 for NCHW44 and NCHW44_DOT(aarch64 binary size grows 2KB) by
2020-06-18 19:40:08 +0800
9cc66963
refactor(sdk/extern-opr): add cpu threads config for MACE by
2020-06-28 17:58:22 +0800
f4694669
fix(dnn/arm): add target os and arch for cross compile by
2020-07-08 16:38:20 +0800
7c278236
feat(core): redact exception messages if logging is disabled by
2020-07-06 15:20:02 +0800
6137adc0
feat(core): move more exceptions to base exception by
2020-07-06 13:22:42 +0800
49f02e71
chore(format): format megbrain/exception.h by
2020-07-06 13:22:42 +0800
9667d93c
feat(core): internal changes to exception types by
2020-07-03 20:17:13 +0800
c8f3742f
fix(mge/doc): fix wrong doc for megengine doc by
2020-07-08 18:01:14 +0800
b8b000db
feat(dnn/fallback): fix fallback interface of weight preprocess by
2020-06-23 19:54:45 +0800
cf3a55ce
fix(mgb/opr-mm): remove PeerDesc from RemoteSend and RemoteRecv by
2020-07-06 18:25:32 +0800
d53dab2f
fix(build): export target megengine by
2020-07-02 16:51:10 +0800