提交 · 4a5e317063cb981ac4eefdab88e12592e82d0405 · deeplearning / MegEngine

13 10月, 2020 6 次提交
- M
  fix(mgb/opr-mm): remove stream -4 for CollectiveComm · 4a5e3170
  由 Megvii Engine Team 提交于 9月 18, 2020
```
GitOrigin-RevId: 41ea88dfa12dd385db122e0777ca18060105839f
```
  4a5e3170
- M
  fix(mgb/core): fix operator input waiting spec · a599725c
  由 Megvii Engine Team 提交于 9月 15, 2020
```
GitOrigin-RevId: 5906dd7ee6c0bdb3c8187681253ee80f50945132
```
  a599725c
- M
  fix(whl/linux): reduce external megengine whl package size · 26e473e7
  由 Megvii Engine Team 提交于 9月 18, 2020
```
GitOrigin-RevId: 3528af8baeab4ed47eee04107665610e9ababdc9
```
  26e473e7
- M
  feat(mgb/jit): add llvm-lit mlir pass check · 5cfb46e1
  由 Megvii Engine Team 提交于 9月 09, 2020
```
GitOrigin-RevId: acc798a49f1fccbc31f03479be3d89f9f30b35e6
```
  5cfb46e1
- M
  fix(mgb): fix optpass fail at transform NCHW to NCHW4 when input dtype is float · 8f44d6ea
  由 Megvii Engine Team 提交于 9月 16, 2020
```
GitOrigin-RevId: 3c2c68b11cbf2dd28ae61745c26cc276e74c0761
```
  8f44d6ea
- M
  feat(mgb/opr): let more ops support empty IO · 95eb6ae3
  由 Megvii Engine Team 提交于 8月 31, 2020
```
GitOrigin-RevId: 84dddb4b23638b29950e438bba2af8b5fd5166fa
```
  95eb6ae3
17 9月, 2020 34 次提交
- M
  fix(build/windows): fix windows build · c974d615
  由 Megvii Engine Team 提交于 9月 17, 2020
```
GitOrigin-RevId: 816fce8bc583ceaaddae43fb0ca851a7c68ea056
```
  c974d615
- M
  fix(mge/build): fix version.ld path · f25d5b56
  由 Megvii Engine Team 提交于 9月 16, 2020
```
GitOrigin-RevId: 6d5bbf0d18df7cd2e0c9b8a42f84730da4873b50
```
  f25d5b56
- M
  fix(mgb/opr): fix VolatileSharedDeviceTensor infinite recursion in make · b7eb4863
  由 Megvii Engine Team 提交于 9月 16, 2020
```
GitOrigin-RevId: f08fa4f327223aa4016835eb6cf8c29d9316ae49
```
  b7eb4863
- M
  fix(mgb/gopt): fix convert batchnorm to elemwise pass issue · fade97d4
  由 Megvii Engine Team 提交于 9月 16, 2020
```
GitOrigin-RevId: eda7f1ab95f9f15448cdcafa2197fa1edd3e7946
```
  fade97d4
- M
  feat(mgb/comp_node): add set_prealloc_config · a7b9ece4
  由 Megvii Engine Team 提交于 9月 03, 2020
```
GitOrigin-RevId: e725e7efdd78e4e8ae85ac963988aef55ee5f9c4
```
  a7b9ece4
- M
  fix(mgb/comp_node): add lock_guard to sd.mtx & update warning message & add test · 066da0bf
  由 Megvii Engine Team 提交于 9月 08, 2020
```
GitOrigin-RevId: ae0bf4b479ceefa59eedaa9bacc9cd56be8dd098
```
  066da0bf
- M
  fix(opr): remove constant value infer for const SharedDeviceTensor · 354b478c
  由 Megvii Engine Team 提交于 9月 09, 2020
```
GitOrigin-RevId: 8fa47b35adbeb8935038b494f6bb2b55df7dd152
```
  354b478c
- M
  feat(opr): add a constant flavor for SharedDeviceTensor · 78d7d400
  由 Megvii Engine Team 提交于 9月 08, 2020
```
Also, add a CONSTANT value inference tag to outputs of
MultipleDeviceTensorHolder.

GitOrigin-RevId: 82a805ed5fed68376c0638a902e303f5a651a478
```
  78d7d400
- M
  fix(test): disable mocked tests on Windows · 14e1a205
  由 Megvii Engine Team 提交于 9月 10, 2020
```
GitOrigin-RevId: 898cfdb09d6b22b2595ecd8db9a3712028928fc6
```
  14e1a205
- M
  fix(mgb/opr): fix ConvBias not passing on prep_filter · 76c79d79
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: 0dc9f9d1331d48f6e8198472332316ef273f80e7
```
  76c79d79
- M
  fix(opr): fix compile on GCC6 · 498f7b2e
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: b59756c923c0fc949041f06c8dc8ab7886bbdd31
```
  498f7b2e
- M
  feat(opr): remove old weight preprocess adapter in fastrun · b9f3077d
  由 Megvii Engine Team 提交于 8月 31, 2020
```
GitOrigin-RevId: f467c0db80b1f917228cbc2e5e9594ee0a00bef5
```
  b9f3077d
- M
  feat(opr): use weight preprocess feature of MegDNN · 75eebb7c
  由 Megvii Engine Team 提交于 7月 20, 2020
```
GitOrigin-RevId: 779041f8a87051e58d5e0ca289773b05a261a8a0
```
  75eebb7c
- M
  refactor(opr): refactor OprArityTrait · fe3ee3cd
  由 Megvii Engine Team 提交于 7月 20, 2020
```
GitOrigin-RevId: fa065cde4ea223dcc394d2a73e14d978384f85de
```
  fe3ee3cd
- M
  fix(dnn/cuda): fix sass loader · 539aac44
  由 Megvii Engine Team 提交于 8月 28, 2020
```
GitOrigin-RevId: f7dec65a2a7eaa24382b6311de3c299cb868bdb8
```
  539aac44
- M
  fix(mgb/atlas): use aclmdlGetOutputSizeByIndex to create DataBuffer · 023d60ae
  由 Megvii Engine Team 提交于 9月 10, 2020
```
GitOrigin-RevId: df7f1b225dd4e39443e4ebf94339f6f47de55102
```
  023d60ae
- M
  build(third_party): update megray · 116eee52
  由 Megvii Engine Team 提交于 9月 11, 2020
```
GitOrigin-RevId: da5e05f82b5112474d51f9eab78318b1d6432742
```
  116eee52
- M
  fix(mgb/jit): mlir doesn't support broadcast · 23437864
  由 Megvii Engine Team 提交于 9月 10, 2020
```
GitOrigin-RevId: 08bfc4c34aa378b61835269a0c0de90d04602bb6
```
  23437864
- M
  feat(mgb/jit): add scalar support for mlir · f87bba68
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: 27b1649c042c68c840a51657807c5ba0ea2e88d8
```
  f87bba68
- M
  fix(mgb/jit): link libdevice.bc when generate nvvm ir · 11b121a7
  由 Megvii Engine Team 提交于 9月 05, 2020
```
GitOrigin-RevId: 49289d65c4d627964c0d53a48f5f911db98012f6
```
  11b121a7
- M
  fix(mgb/jit): fix gpu kernel args order · aa7f28b8
  由 Megvii Engine Team 提交于 8月 27, 2020
```
GitOrigin-RevId: a7f0e56747d43812b71cc8a859b52465e3832c61
```
  aa7f28b8
- M
  feat(mgb/jit): refactor code and add more elemwise mode · 9767ca8f
  由 Megvii Engine Team 提交于 8月 27, 2020
```
GitOrigin-RevId: eb6bcadf54d3111a94152c10d0fb826ca2a16d66
```
  9767ca8f
- M
  feat(mge): add mlir cmake · f9109bed
  由 Megvii Engine Team 提交于 8月 14, 2020
```
GitOrigin-RevId: e86db34b18d75c85f7ef120481f79359dc3e1da2
```
  f9109bed
- M
  feat(mgb/jit): add mlir backend for cpu and cuda · a51d5b4c
  由 Megvii Engine Team 提交于 7月 22, 2020
```
GitOrigin-RevId: 814fed047eb49c9c81e1cedb63b1e3f9b6e7cc27
```
  a51d5b4c
- M
  fix(mgb/dtype): enable TypeCvt for bool when trace(symbolic=True) · dd39265e
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: 4e0fc63369b623e6e9e9eca396ec03f87b56452f
```
  dd39265e
- M
  refactor(mge/profiler): refactor profiler · 6f581906
  由 Megvii Engine Team 提交于 9月 04, 2020
```
GitOrigin-RevId: 279aa779a69e503e38e91ce06a973028824eb0aa
```
  6f581906
- M
  fix(rocm): fix rocm megdnntest sleep and a cut code · cc952b2b
  由 Megvii Engine Team 提交于 9月 04, 2020
```
GitOrigin-RevId: 26de5ca98b28bf93da5c3e9f30c0197eda6d9077
```
  cc952b2b
- M
  fix(mge/parampacksplit): fix param pack split mem forward · a66d4b8b
  由 Megvii Engine Team 提交于 9月 09, 2020
```
GitOrigin-RevId: 8c001b73ffbd086f0cfff7cac2a4c1037bfcecfb
```
  a66d4b8b
- M
  feat(dnn/cuda): add cutlass nchw4 convolution · 76fa7157
  由 Megvii Engine Team 提交于 8月 24, 2020
```
GitOrigin-RevId: 93c9b212f4543e50ec56b5623e1b77bf7209a45b
```
  76fa7157
- M
  fix(mgb/atlas): remove unnessary setdevice · a558d4a2
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: 7492de1f017c676dc74c79b55b13f3428bec7a51
```
  a558d4a2
- M
  fix(mge): fix sublinear memory in jit.trace · 1fe8a212
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: 190a330a8ce2016d24faed0246cc0c294c755946
```
  1fe8a212
- M
  fix(dnn): fix nchw88 winograd weight preprocess · 651920c7
  由 Megvii Engine Team 提交于 9月 07, 2020
```
GitOrigin-RevId: 03c0c52abaa870594e17472814a81d6286cdabf8
```
  651920c7
- M
  fix(whl/mgb/imperative): fix symbols conflict runtime crash · d06f248d
  由 Megvii Engine Team 提交于 9月 05, 2020
```
GitOrigin-RevId: d60c4052c1ac1c4f2b9879c60069d232c8efcd13
```
  d06f248d
- M
  feat(mge/imperative): add sublinear options · 0e82b959
  由 Megvii Engine Team 提交于 9月 01, 2020
```
GitOrigin-RevId: f0e917f716c0bcb700559bb3a11cc5aec97dc117
```
  0e82b959

deeplearning / MegEngine 与 Fork 源项目一致

deeplearning / MegEngine
与 Fork 源项目一致