wangwei
/
MegEngine
Not watched
1
0
Fork 0
Code
Releases 31 Wiki Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain

Commit Graph

  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • | *
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • c2293815 fix(autodiff): proxy_graph_detail::make_backward_graph support multiple opnodes by Megvii Engine Team 2022-03-17 17:20:21 +0800
  • 335d51b4 ci(mge): set env when downloading mkl by huangxinda 2022-03-25 13:04:01 +0800
  • 0948f587 ci(mge): download mkl from ftp by Megvii Engine Team 2022-03-23 15:56:39 +0800
  • 22be7e11 docs(mge/functional): typo by Megvii Engine Team 2022-03-22 08:57:13 +0000
  • 111c150d docs(mge/functional): fix typo by Megvii Engine Team 2022-03-22 08:53:46 +0000
  • 240a685f feat(opencl): opt lite-OpenCL api: opencl_clear_global_data and enable_opencl_deploy lite api by Megvii Engine Team 2022-03-11 18:38:13 +0800
  • 4f60fbbb fix(mge/distributed): add polling to solve xmlrpc server io error by Megvii Engine Team 2022-03-09 17:23:15 +0800
  • 273c0e87 fix(autodiff): fix some bugs in relation to 2nd order grad by Megvii Engine Team 2022-03-01 15:16:53 +0800
  • bc9aa47a feat(mge/indexing): support newaxis by Megvii Engine Team 2022-02-28 13:57:52 +0800
  • 9779bc7f fix(imperative): allow rng op infer shape fallible by Megvii Engine Team 2022-03-02 11:20:58 +0800
  • 8f7fa90c fix(lite): fix lar cuda option crash for mdl model by Megvii Engine Team 2022-03-17 13:52:53 +0800
  • 1071f2ab build(flatbuffer): fix uclibc build flatbuffer by zhanghaolong 2022-03-22 16:09:00 +0800
  • 7582157a chore(deps): change flatbuffers repo from google to MegEngine by zhanghaolong 2022-03-22 14:42:32 +0800
  • 8fa9a8de fix(imperative): fix dot-op implement by Megvii Engine Team 2022-03-19 22:05:07 +0800
  • 6c413ba9 refactor(mge): refactor physical tensor by Megvii Engine Team 2022-03-18 18:33:31 +0800
  • d56570d9 fix(megbrain): add rdnn to copybara by Megvii Engine Team 2022-03-18 18:03:18 +0800
  • 7de1bb11 fix(mge/utils): disable memory forwarding for subgraph by Megvii Engine Team 2022-03-18 16:20:48 +0800
  • b7c9361f perf(mge/functional): add infer_output_attrs_fallible for some ops by Megvii Engine Team 2022-03-17 22:29:30 +0800
  • a4327c4d perf(imperative): add dim_expansion transform for conv/bn1d by Megvii Engine Team 2022-03-17 14:13:11 +0800
  • 72a70dd6 perf(imperative): specialize convolution implementation by Megvii Engine Team 2022-03-14 20:17:32 +0800
  • 12a3ef8d refactor(fastrun): decouple fastrun from computing graph by Megvii Engine Team 2022-03-09 19:07:19 +0800
  • 0a6f4a88 fix(mge/dtr): fix dtr problem by Megvii Engine Team 2022-03-15 14:33:02 +0800
  • 529b394f fix(imperative): fix profiler problem by Megvii Engine Team 2022-03-16 21:24:48 +0800
  • e64536a3 fix(imperative): fix the dtype promote problem when amp by Megvii Engine Team 2022-03-11 18:30:32 +0800
  • 2b80806f perf(imperative/src): improve dot performance by Megvii Engine Team 2022-03-08 20:07:12 +0800
  • 2f3bc2db perf(mge/utils): move astensor1d into C++ by Megvii Engine Team 2022-03-11 13:50:58 +0800
  • fa62f6c0 perf(mge/utils): move convert_input into C++ by Megvii Engine Team 2022-03-10 18:18:50 +0800
  • d98be080 perf(mge): move Const into C++ by Megvii Engine Team 2022-03-09 19:17:43 +0800
  • 1709b394 perf(mge/functional): speed up Broadcast and Reshape by Megvii Engine Team 2022-03-04 16:32:36 +0800
  • 0f736a0a perf(mge/functional): speed up Dimshuffle by Megvii Engine Team 2022-03-04 17:28:54 +0800
  • 3e5e08b0 perf(mge/functional): speed up RemoveAxis by Megvii Engine Team 2022-03-03 19:29:39 +0800
  • a4d473c9 perf(mge/functional): speed up AddAxis by Megvii Engine Team 2022-03-03 19:00:53 +0800
  • 3e206d89 perf(mge/functional): speed up Split by Megvii Engine Team 2022-03-03 17:38:47 +0800
  • 730ddc2d perf(interpreter): improve interpreter performance by Megvii Engine Team 2022-03-09 17:15:32 +0800
  • 729242f9 refactor(imperative): move typecvt code of sereval ops to c++ by Megvii Engine Team 2022-03-07 15:47:59 +0800
  • 3c3fc6f3 refactor(imperative): move python code of elemwise/reduce/conv2d/bn to c++ by Megvii Engine Team 2022-02-23 17:06:21 +0800
  • 84466261 perf(imperative/src): improve elemwise by Megvii Engine Team 2022-03-03 15:51:29 +0800
  • e400b7ff perf(imperative): enable memory forwarding for imperative by Megvii Engine Team 2022-02-09 16:10:46 +0800
  • 84d1a440 fix(imperative): do not use output_desc in rng ops by Megvii Engine Team 2022-03-03 15:02:56 +0800
  • 1ce78aa0 fix(imperative): destruct dnn handles at last by Megvii Engine Team 2022-02-16 14:17:28 +0800
  • 0cb60d64 feat(imperative): add output_descs for apply_on_physical_tensor by Megvii Engine Team 2022-02-23 10:32:02 +0800
  • c7ded2fe refactor(imperative): remove unnecessary reverve in small vector by Megvii Engine Team 2022-02-17 10:42:27 +0800
  • 8c2b916e refactor(imperative): remove some methods in proxy graph by Megvii Engine Team 2022-02-15 16:32:58 +0800
  • 2348a963 refactor(imperative): apply workspace limit hook to mini graph by Megvii Engine Team 2022-02-16 21:36:05 +0800
  • fea46ea9 perf(imperative): add opr cache for apply_on_physical_tensor by Megvii Engine Team 2021-05-11 19:33:02 +0800
  • ea4e6ab9 fix(mgb/opr): fix shape cache of NvOF by Megvii Engine Team 2021-07-01 18:05:12 +0800
  • 3228fb75 fix(cuda): conv algo heuristic choose by Megvii Engine Team 2022-03-17 20:56:34 +0800
  • 8c415f4e feat(dnn): cuda nhwc nearest resize support not 1 or 3 channel by Megvii Engine Team 2022-03-15 20:41:58 +0800
  • 04475744 feat(opencl): add OpenCL cache compat level api by Megvii Engine Team 2022-03-08 13:22:42 +0800
  • 6fb5a343 build(flatbuffer/cx2): fix cx2 build and fix uclibc build flatbuffer by Megvii Engine Team 2022-03-16 18:52:33 +0800
  • 87de704a feat(gopt): fuse conv h_swish by Megvii Engine Team 2022-03-02 19:53:38 +0800
  • 4adba378 feat(lite): add example script and some small change for lar by Megvii Engine Team 2022-03-03 15:27:37 +0800
  • 87f00232 fix(mge/gm): fix missing dtype checking while attach tensors by Megvii Engine Team 2022-03-07 16:24:08 +0800
  • 3726f5cc feat(gopt): merger consecutive relayout and dimshuffle to one relayout to optimize CD4 performarce by Megvii Engine Team 2022-02-25 11:21:51 +0800
  • 1fead9b6 feat(gopt): merge consecutive dimshuffle and relayout to one relayout to optimize CD4 performace by Megvii Engine Team 2022-02-22 19:36:04 +0800
  • 26d1e4f7 feat(gopt): optimize cd4 pass rule for elemwise and typecvt to let cd4 start as soon as possible by Megvii Engine Team 2022-02-22 16:18:24 +0800
  • ac26bdce fix(cuda): fix direct conv speed and memory problem by Megvii Engine Team 2022-03-07 20:01:06 +0800
  • f7994683 feat(cuda): add large kernel direct conv to heuristic algo chooser by Megvii Engine Team 2022-03-07 20:00:14 +0800
  • 6dc0c0b9 fix(dnn): fix the sync problem in some kernels by Megvii Engine Team 2022-03-10 11:50:09 +0800
  • 04193e3b feat(dnn): add nearest mode for remap and resize by Megvii Engine Team 2022-03-01 14:43:04 +0800
  • 69b89388 docs(mge/functional): fix debug_param set_execution_strategy docstring by Megvii Engine Team 2022-03-04 17:09:13 +0800
  • 93c7e451 feat(arm): delete the reduant implement by Megvii Engine Team 2022-02-23 15:41:38 +0800
  • e34a642b feat(fallback): reduce support general intrinsic by Megvii Engine Team 2022-02-17 18:27:26 +0800
  • 10f23778 feat(fallback): add simd general intrinsic by Megvii Engine Team 2022-02-17 18:26:45 +0800
  • 286051ed feat(dnn): differentiate sass kernel with cuda version by Megvii Engine Team 2022-03-01 19:21:10 +0800
  • f78b60ec feat(bazel): make bazel gensass depend on cuda toolchain version automatically by Megvii Engine Team 2021-11-24 19:19:05 +0800
  • f48227c0 feat(mgb): show more details for cuda driver api call by Megvii Engine Team 2021-11-17 20:11:00 +0800
  • bb5af9b4 feat(lite): hidden lar gflags symbols for static link by Megvii Engine Team 2022-02-25 11:05:48 +0800
  • d8bb3ff5 fix(cuda): fix fp16 tensorcore gemm split k workspace by Megvii Engine Team 2022-03-02 10:40:05 +0800
  • 597efed4 feat(lite): add get last error code interface in lite c by Megvii Engine Team 2022-02-25 14:51:27 +0800
  • 90c8a58c docs(docstring): add pad docstring by Megvii Engine Team 2022-02-24 22:50:43 +0800
  • 5f4501e0 fix(gopt): fix conv bias fuse 2 noline by Megvii Engine Team 2022-02-25 21:37:27 +0800
  • ac2f548c docs(imperative/dataloader): update preload description by Megvii Engine Team 2022-02-24 13:25:37 +0800
  • 73b518b7 feat(lite): add get physic addr interface in lite by Megvii Engine Team 2022-02-22 19:00:07 +0800
  • f67086ad fix(lite): fix lite global layout transform symvar replace error by Megvii Engine Team 2022-02-28 10:56:33 +0800
  • c42ce937 (tag: v1.8.2, release-1.8) feat(mge/third_party): update cutlass version by megvii-mge 2022-02-25 11:20:54 +0800
  • 9902ccfc chore(release): bump version by 温娟 2022-02-27 07:07:24 +0000
  • 8e5410e4 feat(cuda): add fp16 compute 16 kernel by Megvii Engine Team 2022-02-23 02:20:14 +0800
  • 472e2f96 refactor(cuda): depthwish large kernel by Megvii Engine Team 2022-02-21 21:54:44 +0800
  • e698ec20 feat(cuda): float16 depthwise large kernel conv compute fp32 by Megvii Engine Team 2022-02-15 22:39:17 +0800
  • 48406382 feat(cuda): support float16 depthwise large kernel conv by Megvii Engine Team 2022-02-07 18:28:06 +0800
  • 7042f76b perf(cuda): speedup conv backward data with small feature map and large filter size by Megvii Engine Team 2022-01-26 19:44:51 +0800
  • 87a2aeeb perf(cuda): speedup chanwise conv with small feature map and large filter size by Megvii Engine Team 2022-01-21 20:29:19 +0800
  • 2293385e feat(mge): add conv padding mode by Megvii Engine Team 2022-01-26 13:27:42 +0800
  • afe9c4b5 feat(dnn/cuda): add implicit bmm kernels for large kernel depthwise convolution backward filter opr by Megvii Engine Team 2022-02-14 16:47:49 +0800
  • e8a16929 feat(dnn/cuda): add heuristic rule for implicit batched gemm large kernel dwconv2d kernels by Megvii Engine Team 2022-02-11 18:07:54 +0800
  • 38067472 fix(dnn/cuda): fix ci by Megvii Engine Team 2022-02-11 17:38:00 +0800
  • 1da58ae1 feat(dnn/cuda): add implicit bmm large kernel dwconv2d dgrad kernels by Megvii Engine Team 2022-02-10 20:24:43 +0800
  • 96050073 feat(dnn/cuda): add implicit bmm large kernel dwconv2d fprop impl by Megvii Engine Team 2022-02-07 18:54:21 +0800
  • 4462953f feat(mge/third_party): update cutlass version by megvii-mge 2022-02-25 11:20:54 +0800
  • d7b0994a feat(cuda): add fp16 compute 16 kernel by Megvii Engine Team 2022-02-23 02:20:14 +0800
  • 8a2e92bd refactor(cuda): depthwish large kernel by Megvii Engine Team 2022-02-21 21:54:44 +0800
  • 6b8a69d5 feat(cuda): float16 depthwise large kernel conv compute fp32 by Megvii Engine Team 2022-02-15 22:39:17 +0800
  • bc385b53 feat(cuda): support float16 depthwise large kernel conv by Megvii Engine Team 2022-02-07 18:28:06 +0800
  • 7d2063e3 perf(cuda): speedup conv backward data with small feature map and large filter size by Megvii Engine Team 2022-01-26 19:44:51 +0800
  • 72403e89 perf(cuda): speedup chanwise conv with small feature map and large filter size by Megvii Engine Team 2022-01-21 20:29:19 +0800
  • 28d48f2f fix(mgb/src): fix megbrain cmake unsupport android_nn by Megvii Engine Team 2022-02-11 12:41:13 +0800
  • ab6d12ca feat(mge): add conv padding mode by Megvii Engine Team 2022-01-26 13:27:42 +0800
  • 177001d5 refactor(dispatch): allow dynamic type creation by Megvii Engine Team 2022-02-16 15:10:11 +0800
  • 150a6a61 perf(dispatch/trace): remove unnecessary h2d for constant by Megvii Engine Team 2022-02-09 03:10:15 +0800