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

Commit Graph

  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • e0fffa52 (tag: v1.2.0, release-1.2) fix(mge/tensor): fix mem leak when getting value in trace by Megvii Engine Team 2021-01-12 18:37:41 +0800
  • 3a4e0598 Revert "perf(mge): add memory optimization for backward graph" by Megvii Engine Team 2021-01-13 10:36:38 +0800
  • 5f299dd0 chore(release): bump version by Megvii Engine Team 2021-01-12 12:14:45 +0800
  • 27503620 fix(fastrun): release free blocks before profiling by Megvii Engine Team 2020-12-25 11:57:24 +0800
  • 08cc1032 fix(imperative): fix persistent_cache by Megvii Engine Team 2021-01-11 10:46:17 +0800
  • 998f71a8 perf(mge/imperative): do not busy wait in imperative's queue by Megvii Engine Team 2021-01-08 11:54:53 +0800
  • 62637fc4 refactor(mgb/utils): allow user set different spin time for ASyncSCQueue by Megvii Engine Team 2021-01-08 11:50:40 +0800
  • d07cfdcb refactor(mgb): move convolution mixin to search_policy by Megvii Engine Team 2021-01-07 02:01:07 +0800
  • a57317f4 refactor(mgb): move profile cache out of mgb opr and update CACHE_KEY_VERSION by Megvii Engine Team 2021-01-06 16:01:55 +0800
  • 278b2baa perf(mge): add memory optimization for backward graph by Megvii Engine Team 2021-01-05 15:02:58 +0800
  • ebe86892 feat(sdk/load_and_run): add flags --io-dump-stdout and --io-dump-stderr by Megvii Engine Team 2021-01-08 16:11:57 +0800
  • 5c7d48cd fix(mge/functional): fix tensor split by Megvii Engine Team 2021-01-04 18:36:27 +0800
  • a240d558 fix(mge/dnn): fix rng and topk oom in distributed training by Megvii Engine Team 2020-12-30 10:46:38 +0800
  • c3c1e46d chore(mgblar): delete winograd transform switch by Megvii Engine Team 2020-12-25 21:52:25 +0800
  • 82847bc7 docs(mge): add conv2d output shape computation formula by Megvii Engine Team 2021-01-09 12:48:15 +0800
  • d3bfb0e9 fix(mge): fix trace exit code and reformat by Megvii Engine Team 2021-01-08 22:11:20 +0800
  • 23b9a98f fix(mge): fix sublnear cuda and mem leak by Megvii Engine Team 2021-01-08 12:56:11 +0800
  • c70a49ed fix(mge): correct trace outputs when grad does copy by Megvii Engine Team 2021-01-08 11:33:11 +0800
  • d4ada69d refactor(mge): trace exception in compiled info by Megvii Engine Team 2021-01-08 10:02:14 +0800
  • c9c3429a refactor(mge): fix sublinear by Megvii Engine Team 2021-01-05 19:11:17 +0800
  • de0742be refactor(mge): reopen passed assertions by Megvii Engine Team 2021-01-03 15:31:00 +0800
  • a90c937d feat(interpreter): add command buffer for inplace by Megvii Engine Team 2020-12-18 14:48:34 +0800
  • 09af925f fix(mge): fix cpp trace function release by Megvii Engine Team 2021-01-05 15:56:16 +0800
  • 3975a54a fix(mge/sublinear): fix divide by zero in sublinear by Megvii Engine Team 2021-01-06 11:49:19 +0800
  • 8f7f52ae feat(jit): add memfwd in jit executor opr by Megvii Engine Team 2020-12-30 21:56:26 +0800
  • dfb2b2ce fix(dnn): change pooling window size smaller than padding constraint to log_error by Megvii Engine Team 2020-12-30 17:57:37 +0800
  • 6919127f fix(mge): fix event pool destructor order issue and do full sync at python exit by Megvii Engine Team 2021-01-05 16:47:06 +0800
  • 60459bd7 fix(mge/pytest): remove sync in pytest teardown & add full_sync by Megvii Engine Team 2021-01-05 16:22:44 +0800
  • c49427d1 feat(imperative): add inplace add_update option in optimizer by Megvii Engine Team 2020-12-02 11:36:43 +0800
  • aeb7980b perf(mgb): outputs of the same opr and same compnode share the same callbackcaller by Megvii Engine Team 2020-12-21 18:26:26 +0800
  • 89b6dbc7 fix(mge/imperative): import subpackages by Megvii Engine Team 2021-01-05 19:20:41 +0800
  • 7f3f9a94 feat(mgb/core): add shape hint for graph optimization by Megvii Engine Team 2020-12-24 18:50:43 +0800
  • d1fbec4f feat(dnn/atlas): add atlas stub by Megvii Engine Team 2020-11-27 14:11:08 +0800
  • a85531dd feat(mgb/opr): add tqt opr by Megvii Engine Team 2020-12-23 14:53:10 +0800
  • 44742e32 fix(mge/api): check input dim of dot and mark output as scalar by Megvii Engine Team 2020-12-28 14:13:23 +0800
  • 697f70c0 feat(mge/pytest): add more tests for specialized grad rules by Megvii Engine Team 2021-01-05 17:10:37 +0800
  • cf3f58cb fix(mge/autodiff): fix segfault when grad is nullptr by Megvii Engine Team 2021-01-04 19:22:57 +0800
  • 288c2e08 fix(mge/autodiff): fix expand_dims and grad rule fallback by Megvii Engine Team 2021-01-04 16:44:44 +0800
  • a5609f3b fix(cambricon): fix cross cn copy for cambricon by Megvii Engine Team 2020-09-03 12:27:47 +0800
  • 05c739b8 refactor(mge/data): rename `MapDataset` to `Dataset` by Megvii Engine Team 2020-12-31 15:24:47 +0800
  • a892e5d0 perf(mge): add more specialized grad rules by Megvii Engine Team 2021-01-02 18:28:00 +0800
  • e9e5f442 fix(mge): expand custom op before trace by Megvii Engine Team 2020-12-31 17:45:12 +0800
  • 3faba54f fix(mge): fix segfault with Function returning unused grads by Megvii Engine Team 2020-12-31 16:30:54 +0800
  • 75129cf0 chore(mge): clean up before merge to dev by Megvii Engine Team 2020-12-31 10:56:40 +0800
  • aba0acc7 fix(sdk): add AssertEqual opr, fix dump_with_testcase_mge by Megvii Engine Team 2020-12-31 10:51:47 +0800
  • dd9f54cd refactor(mge): migrate to new core implementation by Megvii Engine Team 2020-12-31 10:37:19 +0800
  • b9762d71 fix(mge): make parampack run with tensor symbolic shape by Megvii Engine Team 2020-12-30 20:02:53 +0800
  • 4d75f691 feat(mge): restore remote send/recv by Megvii Engine Team 2020-12-30 18:25:10 +0800
  • 9c92701f feat(mge): support python -m megengine.distributed.server by Megvii Engine Team 2020-12-30 18:29:22 +0800
  • 556e0222 refactor(mge/trace): remove apply on varnode by Megvii Engine Team 2020-12-29 15:26:46 +0800
  • 243a05b4 fix(mge): fix cpp trace function release by Megvii Engine Team 2020-12-29 14:10:35 +0800
  • 9fb5581f refactor(mge): add specialized grad rule support by Megvii Engine Team 2020-12-29 12:58:09 +0800
  • 645fc6f0 test(mge/autodiff): add test for weird error by Megvii Engine Team 2020-12-21 23:58:15 +0800
  • 60c7d62a refactor(imperative): remove multidispatch, raw_tensor, register by Megvii Engine Team 2020-12-25 19:40:45 +0800
  • b5e46ae9 feat(mge): restore Function by Megvii Engine Team 2020-12-25 14:38:47 +0800
  • dc250745 feat(mge): add python custom op by Megvii Engine Team 2020-12-25 14:03:53 +0800
  • 60c44b08 refactor(mge): refactor to prepare for custom grad rules by Megvii Engine Team 2020-12-22 17:35:00 +0800
  • 61f65cd4 test(mge): fix megbrain_graph/cgtools test by Megvii Engine Team 2020-12-28 21:38:25 +0800
  • ae47fd4e fix(mge): fix none return value for attrs, add test_correctness by Megvii Engine Team 2020-12-22 14:48:32 +0800
  • 97d12b3e refactor(mge/imperative): fix compilation for Python 3.5 by Megvii Engine Team 2020-12-19 16:28:30 +0800
  • 87f4b46e perf(mge/imperative): move convert_inputs from python to C++ by Megvii Engine Team 2020-12-21 22:18:10 +0800
  • b310f261 feat(mge/imperative): implement trace and dump under new core implementation by Megvii Engine Team 2020-12-10 22:47:57 +0800
  • 14d8b709 perf(mge/imperative): add mini graph to partially replace proxy graph by Megvii Engine Team 2020-12-18 12:50:50 +0800
  • c294b9d1 refactor(mge/tensor): remove old implementation by Megvii Engine Team 2020-12-18 18:51:14 +0800
  • 15e8e7df refactor(mge/imperative): migrate to new core implementation by Megvii Engine Team 2020-12-17 18:30:10 +0800
  • 34c705fc refactor(mge/imperative): move detach into C++ by Megvii Engine Team 2020-12-15 18:24:27 +0800
  • 147cef52 refactor(mge/imperative): implement new tensor system by Megvii Engine Team 2020-12-01 14:12:15 +0800
  • 7f48625f chore(third_party): add range-v3 as third_party dependency by Megvii Engine Team 2020-12-14 20:29:54 +0800
  • 184e1311 fix(mgb/opr): move NVOF opr's shape inference to execute part by Megvii Engine Team 2020-12-26 17:31:41 +0800
  • d168cea4 feat(opr): add param(axis) for GetVarShape by Megvii Engine Team 2020-12-14 14:03:29 +0800
  • e9c036cc fix(mge/imperative): fix op specializations and tuple hash by Megvii Engine Team 2020-12-19 14:12:07 +0800
  • b5ec83c5 test(mge/autodiff): validate grad is set to None when no dependency by Megvii Engine Team 2020-12-21 15:36:26 +0800
  • 577366a2 feat(mge/imperative): basic impl of python c extension for opdef by Megvii Engine Team 2020-12-04 22:35:34 +0800
  • 9d928e7f refactor(mge/distributed): sync interpreter for distribtued launcher by Megvii Engine Team 2020-12-26 13:37:16 +0800
  • 4e9be159 feat(mgb/gopt): add opt pass for fusing convolution and reformat by Megvii Engine Team 2020-11-24 17:03:09 +0800
  • c3a4b222 feat(dnn/cuda): add cutlass impls for fused convolution reformat operation by Megvii Engine Team 2020-11-24 16:59:59 +0800
  • 5f44203d feat(dnn/cuda): add a cutlass impl for fusing convolution and dimshuffle by Megvii Engine Team 2020-11-17 12:47:13 +0800
  • 61f917fb feat(dnn/cuda): add impl for fusing warp perspective and dimshuffle by Megvii Engine Team 2020-11-21 18:33:11 +0800
  • 15dd5e1a fix(mgb/core): fix memory management release cambricon var issue by Megvii Engine Team 2020-11-04 16:36:12 +0800
  • 38ea5f1b fix(mgb/jit): lower FuncOp to GPUFuncOp directly by Megvii Engine Team 2020-12-22 17:05:11 +0800
  • 4cfedc16 build(imperative): clean develop before build imperative and mgb by Megvii Engine Team 2020-12-15 18:41:21 +0800
  • 9a79342d refactor(mge/virtualdep): remove virtualdep opdef by Megvii Engine Team 2020-12-14 18:39:49 +0800
  • 88b402ef feat(mge/trace): tracing use id to set priority by Megvii Engine Team 2020-12-11 11:53:51 +0800
  • d2910f7e fix(mgb/jit): add bind_shape feature to MLIRCompiler by Megvii Engine Team 2020-12-19 15:34:55 +0800
  • 9de1ea6a perf(imperative): add apply_on_physical_tensor for Elemwise by Megvii Engine Team 2020-12-14 13:56:44 +0800
  • 469d0808 fix(asan/thread): fix asan issue by Megvii Engine Team 2020-12-16 17:09:59 +0800
  • eb826422 fix(dnn): forbid pooling window size smaller than padding by Megvii Engine Team 2020-12-15 18:47:23 +0800
  • a78c1109 fix(imperative): add param(axis) for GetVarShape by Megvii Engine Team 2020-12-14 14:03:29 +0800
  • cde9727a fix(mge/quantization): fix `qat.conv_bn` parameter issue by Megvii Engine Team 2020-12-17 10:39:49 +0800
  • 94796060 test(mge/quantization): qat module, quantized op and module by Megvii Engine Team 2020-12-15 14:41:05 +0800
  • ab9f44f1 feat(mge/quantization): add support for easyquant by Megvii Engine Team 2020-11-19 15:18:14 +0800
  • fc0fcd2f chore(winograd): remove winograd transform code by Megvii Engine Team 2020-11-13 12:39:30 +0800
  • d1adc9a2 fix(dnn): fix opencl algo search by Megvii Engine Team 2020-12-16 00:22:40 +0800
  • 368c1860 fix(mgb/jit): find cuda include path correctly by Megvii Engine Team 2020-11-03 21:13:09 +0800
  • b04e0466 feat(megbrain): add alias name to model serialization by Megvii Engine Team 2020-12-09 18:20:41 +0800
  • cf53d9e0 fix(mgb/tensor): do tensor overlap check only when d2d and h2h by Megvii Engine Team 2020-12-07 19:04:35 +0800
  • 7e2b2dbf fix(dnn/test): delete large size in ARM_COMMON.FP32_GEVM by Megvii Engine Team 2020-12-14 10:57:41 +0800
  • 69e3e322 feat(imperative): auto generated opdef header and python binding by Megvii Engine Team 2020-12-18 10:12:14 +0800
  • 0398a786 fix(build/windows/cuda/llvm): fix windows bazel build with cuda * Adapt to the new version llvm/clang-11 * fix windows bazel build with cuda * add windows bazel build cuda ci * opt windows bazel ci scripts by Megvii Engine Team 2020-12-11 08:22:44 +0800
  • b9c37112 refactor(mge/distributed): skip barrier when running with single node by Megvii Engine Team 2020-12-11 14:53:38 +0800