Page History

Commits since 1.7.0 release:

...

Features

CUDA Graphs

a35d568a0 - GitHub,

...

Sat Sep

...

19 15:

...

17:

...

12 2020 -0700 :

...

[1.x] Enable CUDA Graphs for TRT (#19184)

0fce3812f - GitHub, Fri Sep

...

18 20:

...

18:

...

50 2020 -0700 :

...

[1.x][FEATURE] CUDA graphs support (#19142)

CUDA 11 Support

e1bcb336f - GitHub,

...

Mon Sep

...

14 18:

...

12:

...

38 2020 -0700 : [

...

1.x]

...

Backport 'Update CUB and include it only for CUDA < 11 #18799' (#18975)

TensorRT

606933f2c - GitHub, Tue Sep 15 15:58:54 2020 -0700 : TensorRT: add int8 with calibration (#19011)

2d077db1c - GitHub, Thu Sep 10 13:12:44 2020 -0700 : Add TRT verbose mode (#19100)

8dcc65304 - GitHub, Wed Aug 19 00:22:02 2020 -0700 : [1.x] Backporting TensorRT-Gluon Partition API (and TensorRT 7 support) (#18916)

OneDNN

49ba44afd - GitHub, Fri Sep 18 10:10:35 2020 +0800 : [v1.x][Submodule] Upgrade to oneDNN v1.6.3 (#19153) (#19161)

171110371 - GitHub, Tue Aug 11 21:32:58 2020 +0800 : [1.x][submodule] Update oneDNN to official v1.6 release (#18867) (#18867)

eae6171cb - GitHub, Tue Aug 4 12:26:46 2020 -0700 : [submodule] Upgrade to oneDNN v1.6 (#18822)

IntGemm

d2e645204 - GitHub, Wed Sep 16 09:41:35 2020 -0700 : [1.x] Backport of intgemm #17559 (#19099)

Subgraph API

9dfac79d5 - GitHub, Mon Sep 14 22:57:05 2020 -0700 : [1.x] Backport Fix for duplicate subgraph inputs/outputs (#16131) (#19112)

Extensions

938328299 - GitHub, Mon Sep 14 10:31:36 2020 -0700 : [1.x] Backport #19103 (#19117)

b5e9c9984 - GitHub, Wed Sep 2 20:40:54 2020 -0700 : [1.x] Backporting #19016 (#19069)

9445a2dc0 - GitHub, Wed Aug 19 18:44:50 2020 -0700 : Backport: Change Partition API's options_map to std::unordered_map #18929 (#18964)

d1ac7c849 - GitHub, Tue Aug 18 15:50:30

...

975aa6e0d - GitHub, Wed Sep 23 00:58:50 2020 -0700 : [WEBSITE] v1.8 website patch (#19212)

c2df97faa - GitHub, Tue Sep 22 09:55:44 2020 -0700 : Tweeking syntax to be closer to other tests (#19186) (#19206)

96f645419 - GitHub, Tue Sep 22 09:55:14 2020 -0700 : [v1.8.x] ElementWiseSum fix for oneDNN (#18777) (#19200)

fe7cf9911 - GitHub, Mon Sep 21 18:03:17 2020 -0700 : Fix flaky intgemm test in v1.8.x too (#19204)

...

2020 -0700 : [1.x]

...

Backporting #18779 to v1.x (#18894)

ONNX

b888d3c71 - GitHub, Fri Sep

...

11 10:17:23 2020 -0700 : [v1.x] Update onnx support to work with onnx 1.7.0 with most CV models (#19017)

Large Tensor

e6de5ae47 - GitHub, Fri Jul 24 09:50:46 2020 -0700 : Fix linalg_potri and linalg_potrf operators for large tensor. (#18752)

d0093458e - GitHub, Mon Jul 27 20:28:43 Revert "Fix memory leaks in Gluon (#18328) (#18359)" (#19181)0fce3812f - GitHub, Fri Sep 18 20:18:50 2020 -0700 : [1.x][FEATURE] CUDA graphs support (#19142)LT] Add forward, backward test for linalg.gemm2 (#18784)

566d9d348 - GitHub, Mon Jul 27 10:04:30 5079c3514 - GitHub, Thu Sep 17 21:57:59 2020 -0700 : [v1.x] Backport Improve environment variable handling in unittests (#18424) (#19173)add large matrix tests for linalg ops: det, inverse, trsm, trmm (#18744)

85ff00d5c - GitHub, Fri Jul 24 15:55:05 2020 -0700 : Add Large Tensor Test for linalg_syrk (#18782)

f4e62dfe9 - GitHub, Thu Jul 30 22:49:36 2020 -0700 : Add Large Dim Checks for linalg Operators (#18816)

84c9e0d46 - GitHub, Thu Jul 30 10:40:32 2020 -0700 49ba44afd - GitHub, Fri Sep 18 10:10:35 2020 +0800 : [v1.x][Submodule] Upgrade to oneDNN v1.6.3 (#19153) (#19161)ce0a518e7 - GitHub, Thu Sep 17 15:47:32 2020 -0700 : [v1.x] Backport Unittest tolerance handling improvements (#18694). Also test seeding (#18762). (#19148)LT] Add forward & backward linalg.gemm test for large size (#18825)

ca6bcf348 - GitHub, Wed Jul 29 15:16:10 2020 -0700 : adding error message when attempting to use Large tensor with linalg_syevd (#18807)

Website Improvements

975aa6e0d - GitHub, Wed Sep 23 00:58:50 2020 -0700 : [WEBSITE] v1.8 website patch (#19212)

Documentation

5051fd9a5 - GitHub, Tue Sep 8 10:49:08 2020 -0700 : Fix mxnet.test_utils.check_numeric_gradient documentation (#19060)

cc287a0fc - GitHub, Fri Aug 7 22:09:24 2020 -0700 : Update windows_setup.md (#18874)

License

d144edd46 - GitHub, Tue Sep 1 11:03:17 2020 -0700 : 1.x: Stop packaging GPL libquadmath.so (#19055)

0b5b44948 - GitHub, Mon Aug 10 12:19:04 2020 -0700 : Remove mention of nightly in pypi (#18635) (#18884)

2aa270288 - GitHub, Mon Aug 3 12:33:39 2020 -0700 : [v1.x] Mkldnn header fix v1x for nightly binaries (#18797)

CI Improvements

620d058f7 - GitHub, Thu Sep 17 09:32:05 2020 -0700 : [v1.x] Add new CI pipeline for building and testing with cuda 11.0. (#19149)

...

9981e847f - GitHub,

...

Tue Aug 18 08:32:35 2020 -0700 : [CI][1.x] Cherrypick: Upgrade unix gpu toolchain (#18186) (#18785)

9fbf3d310 - GitHub, Wed Aug 12 10:12:06 2020 -0700 : Fix CI in v1.x branch (#18907)

Bug Fixes

07d7e139d - GitHub, Fri Sep 25 18:03:40 2020 -0700 : delete executor before reallocating it memory (#19222)

7c9046a6d the error of gradient of np.pad (#19044) (#19167)3d9af6e32 - GitHub, Wed Sep 16 23 22:1533:15 47 2020 -0700 : [1v1.x] Backport Add cmake flag USE_FATBIN_COMPRESSION, ON by default (#19123) (#19158)Nightly Large Tensor test cherrypicks (#19194) (#19215)

c2df97faa 837c7e43c - GitHub, Wed Tue Sep 16 1722 09:3055:06 44 2020 -0700 : SymbolBlock.imports ignore_extra & allow_missing (#19156)Tweeking syntax to be closer to other tests (#19186) (#19206)

96f645419 d2e645204 - GitHub, Wed Tue Sep 16 22 09:4155:35 14 2020 -0700 : [1v1.8.x] Backport of intgemm #17559 (#19099)ElementWiseSum fix for oneDNN (#18777) (#19200)

fe7cf9911 606933f2c - GitHub, Tue Mon Sep 15 1521 18:5803:54 17 2020 -0700 : TensorRT: add int8 with calibration (#19011)Fix flaky intgemm test in v1.8.x too (#19204)

04966902e 9dfac79d5 - GitHub, Mon Fri Sep 14 18 22:5717:05 50 2020 -0700 : [1.x] Backport Fix for duplicate subgraph inputs/outputs (#16131) (#19112)Revert "Fix memory leaks in Gluon (#18328) (#18359)" (#19181)

5079c3514 - GitHub, Thu Sep 17 21:57:59 e1bcb336f - GitHub, Mon Sep 14 18:12:38 2020 -0700 : [1v1.x] Backport 'Update CUB and include it only for CUDA < 11 #18799' (#18975)Improve environment variable handling in unittests (#18424) (#19173)

ce0a518e7 938328299 - GitHub, Mon Thu Sep 14 1017 15:3147:36 32 2020 -0700 : [1v1.x] Backport #19103 (#19117)Unittest tolerance handling improvements (#18694). Also test seeding (#18762). (#19148)

039fef9f1 8b5687467 - GitHub, Sat Thu Sep 12 1817 08:5852:31 08 2020 -0700 : Fix race condition in NaiveEngine::PushAsync (#19108) (#19122)the error of gradient of np.pad (#19044) (#19167)

3d9af6e32 b888d3c71 - GitHub, Fri Wed Sep 11 1016 23:1715:23 15 2020 -0700 : [v1.x] Update onnx support to work with onnx 1.7.0 with most CV models (#19017)1.x] Backport Add cmake flag USE_FATBIN_COMPRESSION, ON by default (#19123) (#19158)

837c7e43c - GitHub, Wed Sep 16 17:30:06 2020 -0700 : SymbolBlock.imports ignore_extra & allow_missing (#19156)

8b5687467 - GitHub, Sat Sep 12 18:58:31 2020 -0700 : Fix race condition in NaiveEngine::PushAsync (#19108) (#191222d077db1c - GitHub, Thu Sep 10 13:12:44 2020 -0700 : Add TRT verbose mode (#19100)

a9bd1d2f1 - GitHub, Tue Sep 8 21:21:51 2020 -0700 : empty list cannot be cleared issue fixed. (#14882)

5051fd9a5 - GitHub, Tue Sep 8 10:49:08 2020 -0700 : Fix mxnet.test_utils.check_numeric_gradient documentation (#19060)

8dbed966e - GitHub, Mon Sep 7 13:35:45 2020 -0700 : Update base_module.py (#19096)

748eebd74 - GitHub, Thu Sep 3 09:21:50 2020 -0700 : fix block.export (#17970) (#19075)

b5e9c9984 - GitHub, Wed Sep 2 20:40:54 2020 -0700 : [1.x] Backporting #19016 (#19069)

1a398e580 - GitHub, Wed Sep 2 18:47:01 2020 -0700 : Support for fp16 in SpM x DnsM on GPU (#18930) (#19074)d144edd46 - GitHub, Tue Sep 1 11:03:17 2020 -0700 : 1.x: Stop packaging GPL libquadmath.so (#19055)

bce4cc69d - GitHub, Tue Aug 25 18:18:31 2020 -0700 : [1.x] Backport of Fix LeakyRelu behaviour on empty input (#18934) (#19009)

dfefe8799 - GitHub, Mon Aug 24 10:00:52 2020 -0700 : Get rid of monkey patching in LossScaler overflow handling (#18959) (#18973)

9445a2dc0 - GitHub, Wed Aug 19 18:44:50 2020 -0700 : Backport: Change Partition API's options_map to std::unordered_map #18929 (#18964)

8dcc65304 - GitHub, Wed Aug 19 00:22:02 2020 -0700 : [1.x] Backporting TensorRT-Gluon Partition API (and TensorRT 7 support) (#18916)

b4da2dd59 - GitHub, Tue Aug 18 21:45:22 2020 -0700 : remove upper bound (#18857) (#18910)d1ac7c849 - GitHub, Tue Aug 18 15:50:30 2020 -0700 : [1.x] Backporting #18779 to v1.x (#18894)

9981e847f - GitHub, Tue Aug 18 08:32:35 2020 -0700 : [CI][1.x] Cherrypick: Upgrade unix gpu toolchain (#18186) (#18785)

6ae469a17 - GitHub, Tue Aug 18 15:10:57 2020 +0800 : fix gelu to use erf based algorithm (#18827) (#18946)

...

d2d6408aa - GitHub, Thu Aug 13 16:44:29 2020 -0700 : bump version to 1.8.0 (#18899)

9fbf3d310 - GitHub, Wed Aug 12 10:12:06 2020 -0700 : Fix CI in v1.x branch (#18907)

171110371 - GitHub, Tue Aug 11 21:32:58 2020 +0800 : [1.x][submodule] Update oneDNN to official v1.6 release (#18867) (#18867)

0b5b44948 - GitHub, Mon Aug 10 12:19:04 2020 -0700 : Remove mention of nightly in pypi (#18635) (#18884)

cc287a0fc - GitHub, Fri Aug 7 22:09:24 2020 -0700 : Update windows_setup.md (#18874)

eae6171cb - GitHub, Tue Aug 4 12:26:46 2020 -0700 : [submodule] Upgrade to oneDNN v1.6 (#18822)

2aa270288 - GitHub, Mon Aug 3 12:33:39 2020 -0700 : [v1.x] Mkldnn header fix v1x for nightly binaries (#18797)

5db8dee9b - GitHub, Mon Aug 3 12:05:20 2020 -0700 : [ONNX export] Fixing spatial export for batchnorm (#17711) (#18846)

...

1a31cea22 - GitHub, Fri Jul 31 13:55:43 2020 -0700 : Add unit tests for potri and potrf backward and check output shape in unit tests. (#18803)

f4e62dfe9 - GitHub, Thu Jul 30 22:49:36 2020 -0700 : Add Large Dim Checks for linalg Operators (#18816)

84c9e0d46 - GitHub, Thu Jul 30 10:40:32 2020 -0700 : [v1.x][LT] Add forward & backward linalg.gemm test for large size (#18825)

ca6bcf348 - GitHub, Wed Jul 29 15:16:10 2020 -0700 : adding error message when attempting to use Large tensor with linalg_syevd (#18807)

85eb528c1 - GitHub, Tue Jul 28 23:08:53 2020 -0700 : Add syrk test shape check (#18812)

7bef9cb23 - GitHub, Tue Jul 28 16:58:07 2020 -0700 : Back port optimization to broadcast_axis to MXNet1.x (#18773)

d0093458e - GitHub, Mon Jul 27 20:28:43 2020 -0700 : [1.x][LT] Add forward, backward test for linalg.gemm2 (#18784)

566d9d348 - GitHub, Mon Jul 27 10:04:30 2020 -0700 : [v1.x] add large matrix tests for linalg ops: det, inverse, trsm, trmm (#18744)