e4d42cd8
feat(imperative): add more tools for megengine by
2021-12-15 19:50:09 +0800
2b8e7940
fix(lite/cambricon): fix cambricon models which have multiple comp node by
2021-12-10 14:51:36 +0800
cfad9a5d
fix(mgb/cambricon): fix magicmind runtime opr when set workspace point second time by
2021-12-03 11:50:18 +0800
262e124b
feat(lite): add lite interface for cambricon models by
2021-11-02 15:23:28 +0800
0c3699ff
test(imperative/python): fix testcase for magicmind runtime module by
2021-11-01 18:27:35 +0800
f398c8e6
feat(imperative): add magicmind runtime opr by
2021-10-29 14:58:08 +0800
f06c890e
fix(dnn/cambricon): make checksum bang-c kernel work for MLU370 by
2021-10-28 13:26:08 +0800
396412e7
test(mgb/cambricon): add testcase for magicmind runtime opr by
2021-10-29 14:40:36 +0800
b5284796
feat(mgb/cambricon): add magicmind runtime opr by
2021-10-26 18:52:21 +0800
e8e535b7
(revert-411-add-tools)
Revert "feat(imperative): add more tools for megengine" by
2021-12-15 19:39:09 +0800
6f74780a
Merge pull request #413 from kagome1007:master by
2021-12-15 11:58:18 +0800
6ce4a344
feat(dnn): add fallback postprocess by
2021-12-14 10:57:53 +0800
be84d13f
fix(lite/load_and_run): fix cuda options build error for load and run by
2021-12-09 18:32:04 +0800
9646f1aa
fix(lite/load_and_run): fix invalid model type caused by oss model magic change by
2021-12-08 12:09:08 +0800
623477fd
refactor(megbrain): combine mc40 and mc20 compnode and change mc20 to ax by
2021-10-29 15:30:25 +0800
11556cba
feat(megbrain): add mc20 bazel by
2021-10-27 21:05:56 +0800
d2c37fe4
feat(megbrain): add mc20 cmake by
2021-10-27 15:25:42 +0800
b8f810ee
feat(megbrain): add mc20 runtime opr by
2021-10-25 19:57:49 +0800
21345d6b
fix(cmake): only add -Werror=unused-parameter at CI env by
2021-12-10 16:10:21 +0800
744efe72
fix(mge/utils): fix module_status import error by
2021-12-08 18:38:37 +0800
667e6099
Merge pull request #411 from MegEngine/add-tools by
2021-12-14 18:45:26 +0800
8b88558a
test(lite): add lite test readme by
2021-12-14 17:37:50 +0800
b3975685
(docstring-reshape)
docs(mge/functional): update functional.tensor.reshape docstring by
2021-12-07 13:41:28 +0800
40fa1b94
feat(imperative): add more tools for megengine by
2021-12-10 16:17:00 +0800
da8b16fc
docs(mge/distributed): change the allreduce related functions document by
2021-12-09 16:58:03 +0800
67e4e834
(tag: v1.7.1.m1)
fix(lite): fix the force_output_use_user_specified_memory when out var not supported by
2021-12-08 11:00:43 +0800
0d37bfb0
test(lite): add ax model test by
2021-11-25 16:11:51 +0800
cb8ebb54
fix(mgb): whl use gcc8 by
2021-12-08 16:14:27 +0800
315b908c
feat(imperative): mge hub pretrained support s3 by
2021-11-21 19:08:57 +0800
c96dbd29
fix(dnn/arm_common): support more monotonous case in arm typecvt for performance by
2021-12-03 16:03:23 +0800
2e441ed0
docs(mge/functional): update functional.clip docstring by
2021-12-07 11:48:04 +0800
02943036
docs(mge/functional): update functional.tensor.reshape docstring by
2021-12-07 13:41:28 +0800
2d1d5e9b
docs(mge/functional): update functional.mul docstring by
2021-12-06 19:21:39 +0800
541bf3af
fix(mge/functional): fix one_hot irregular coding style by
2021-12-08 12:34:25 +0800
409aa489
docs(mge/functional): update functional.ones docstring by
2021-12-06 18:49:54 +0800
695d24f2
ci(scripts): add JetsonNano cu102 build option by
2021-12-02 11:37:13 +0800
3d6fe33e
Revert "fix(imperative): check async error when getting value" by
2021-12-02 16:30:40 +0800
7e2d8912
ci(docstring): add sphinx autodoc build test for docstring by
2021-11-05 19:02:05 +0800
ead611e1
(tag: v1.7.0.m1)
perf(dnn): slightly improve arm neon transcendental function performance by
2021-11-30 17:48:18 +0800
96ffcb2d
docs(mge/functional): update functional.batch_norm docstring by
2021-12-02 19:28:01 +0800
0d169524
fix(mgb/cuda): fix conv error when the input tensor is too large by
2021-11-25 14:05:00 +0800
4f86b2c5
(tag: v1.7.0, release-1.7)
fix(mge/subgraph): fix mem leak by
2021-11-19 10:15:50 +0800
cbf9f136
fix(imperative): fix CompNodeSyncManager deconstruct by
2021-11-26 21:01:46 +0800
618c77c0
feat(dnn): enable eye to support bool by
2021-11-03 13:52:44 +0800
a8941af0
fix(docs): fix traced module docs warning by
2021-11-29 17:49:23 +0800
fefd4f09
fix(traced_module): fix tensor node init by
2021-11-29 18:23:41 +0800
ee634beb
fix(pylite): fix device_id to dev_id by
2021-11-30 19:11:00 +0800
8f2e2711
fix(pylite): add dtype to empty tensor to_numpy api by
2021-11-27 11:58:09 +0800
02d5f46d
fix(mgb/x86): fix convbias crash on X86 by
2021-11-30 11:46:18 +0800
accb2d8d
fix(mgb/serialize): fix flatbuffer compatibility issues by
2021-09-29 17:15:55 +0800
5b1383e0
fix(mgb/plugin): fix io dump for qint4, quint4 and bool type tensors by
2021-11-22 17:51:50 +0800
5e07e1e0
fix(dnn/falback): let cpu be able to execute int4 model by
2021-11-22 17:50:56 +0800
0c098a8c
fix(lite/lar): fix input data parse for lar by
2021-11-22 17:47:23 +0800
2696e4ef
feat(dnn): add float16 for remap backward by
2021-10-29 15:40:12 +0800
ff171934
fix(imperative): fix CompNodeSyncManager deconstruct by
2021-11-26 21:01:46 +0800
2065f0c2
fix(mge/subgraph): fix mem leak by
2021-11-19 10:15:50 +0800
57e19747
fix(imperative): check async error when getting value by
2021-11-17 12:38:49 +0800
1f0cc891
feat(dnn): enable eye to support bool by
2021-11-03 13:52:44 +0800
d3c7c563
fix(traced_module): fix tensor node init by
2021-11-29 18:23:41 +0800
e9a4cffe
fix(docs): fix traced module docs warning by
2021-11-29 17:49:23 +0800
b22ceba7
docs(imperative): fix code exmaple indent warnings by
2021-11-25 13:52:58 +0800
7ac693a8
docs(imperative): fix multiple target warning when cross-reference by
2021-11-25 10:24:36 +0800
e65e3f05
fix(lite/load_and_run): fix some bugs of load and run by
2021-11-29 10:48:26 +0800
24f12df9
chore(release): bump version by
2021-11-29 16:17:43 +0800
605836a7
fix(mgb): fix trt5.1 test can not create FlattenConcatPlugin by
2021-11-27 00:29:51 +0800
11d75fec
feat(dnn/check_non_finite): add batch check_non_finite by
2021-10-28 11:18:55 +0800
7a023c05
feat(mge/traced_module): add optimization api by
2021-11-26 15:45:23 +0800
6c692b26
feat(mge/traced_module): add some fuse passes by
2021-11-26 15:44:11 +0800
b28ad4e8
feat(mge/traced_module): add pattern match for TracedModule by
2021-11-23 19:48:10 +0800
2318ea3f
fix(dnn): fix naive average pooling overflow bug for int8 type by
2021-11-19 19:40:14 +0800
f64a2e02
Revert "fix(mge/subgraph): fix mem leak" by
2021-11-26 17:13:04 +0800
ea3eb66e
docs(mge): modify some docstring about conv and tools by
2021-11-25 13:10:31 +0800
acd835d0
docs(mge): add some docs for cvt_color and GraphInference by
2021-11-23 20:48:12 +0800
0185c1a9
feat(mge/traced_module): add argspec for top TracedModule by
2021-11-26 11:28:57 +0800
1daeba76
fix(mge/traced_module): fix bn opdef loader by
2021-11-25 10:47:02 +0800
2d20f937
fix(mge/traced_module): fix merge GetAttr failure when qualname is incorrect by
2021-11-15 14:02:29 +0800
e6c271ae
fix(mge/traced_module): fix some bugs for graph surgery by
2021-11-24 13:29:29 +0800
2d54ad18
feat(lite): add global layout transform interface for load and run by
2021-11-12 11:18:57 +0800
ba2f0c2e
fix(dnn/cuda): fix cudnn_conv algo of conv_bias opr for fp16 add z cases by
2021-11-16 19:10:47 +0800
30976c23
fix(mgb/gopt): fix global layout transform by
2021-10-20 17:37:58 +0800
b6c0e8d0
fix(mgb/gopt): update profiler cache of global layout transform test by
2021-10-19 13:58:12 +0800
f2db7b0d
feat(mgb/gopt): global layout transform support cuda fp16 by
2021-09-24 16:54:22 +0800
ca7cec7a
fix(mgb/gopt): minor fixes for global layout transform by
2021-10-18 15:47:11 +0800
93152dfa
fix(mgb/gopt): fix global layout transform by
2021-10-18 14:17:52 +0800
fe93013a
feat(mgb/gopt): global layout transform support nchw_nchwxx hybrid mode by
2021-09-22 18:23:52 +0800
3d45d352
feat(mgb/gopt): profiler support checking algo availability by
2021-09-24 14:50:18 +0800
5f15f759
test(mgb/gopt): add a testcase for SubGraphExtractor with multiple outputs by
2021-10-20 17:52:02 +0800
a6230ba9
feat(mgb/gopt): global layout transform support arm by
2021-09-24 14:54:41 +0800
0be6ca88
fix(src/core): fix record change ptr bug on comp node copy by
2021-11-25 16:26:41 +0800
84baf3df
feat(mgb): add tensorrt plugin support by
2021-10-28 20:58:39 +0800
b59e8ccf
fix(mgb): fix cambricon bangc copybara by
2021-11-17 14:19:34 +0800
3116e128
fix(ci/integration_test): fix benchmark torch version by
2021-11-17 23:31:27 +0800
66f13586
fix(mge/network): add warning when the node is not in the output list by
2021-11-24 14:35:00 +0800
fe6299c5
Revert "feat(mgb): allow output tensor's ptr change when record" by
2021-11-24 14:20:19 +0800
c85631aa
feat(dnn): use ref ptr interface for all backends by
2021-11-24 14:16:16 +0800
d90cb776
feat(src/core): record support change ptr basic by
2021-11-24 14:15:47 +0800
7b19bc76
feat(traced_module): support traced module backward compatible serialization by
2021-09-08 18:20:32 +0800
ffbfe59c
fix(mge/subgraph): fix mem leak by
2021-11-19 10:15:50 +0800
89186edc
fix(dnn): correct reduce/argmxx/fakequant calculation with nan by
2021-11-02 17:38:13 +0800
68cdabd2
feat(opr): indexing_multi_axis_vec support nd index by
2021-10-13 17:30:05 +0800