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

Commit Graph

  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • | *
  • * |
  • * |
  • * |
  • * |
  • * |
  • |\ \
  • | * |
  • |/ /
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • | | *
  • | |/
  • | | *
  • | |/
  • * |
  • * |
  • * |
  • * |
  • * |
  • |/
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • |\
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • | | *
  • | | *
  • | | *
  • | * |
  • * | |
  • * | |
  • * | |
  • * | |
  • * | |
  • * | |
  • * | |
  • * | |
  • | | *
  • | |/
  • |/|
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • 0f1afb09 feat(fallback): imp gi matmul AlgoF32GiMK4_4x8 algo, move AlgoF32GemvMK4 from arm_common to fallback by Megvii Engine Team 2022-05-06 15:00:53 +0800
  • 410dcb6c feat(fallback): add more gi api for conv, and add gi API test by Megvii Engine Team 2022-05-06 14:54:03 +0800
  • a0e53118 fix(src/comp_node): fix calling cuda driver api by Megvii Engine Team 2022-04-21 15:44:55 +0800
  • ccea0e23 fix(dnn/rdnn): add warmup before profile by Megvii Engine Team 2022-04-27 19:09:05 +0800
  • 58df717e fix(mge/autodiff): fix attaching tensor already in gradient path by Megvii Engine Team 2022-04-29 14:39:17 +0800
  • 05186e7b fix(midout): fix elemwise crash after midout some dnn backends opr will use agency opr, for example: softmax cpu naive imp will call elemwise opr, at model dump stage, we can not get dnn runtime logic, so we record elemwise mode info at runtime stage. by Megvii Engine Team 2022-05-06 19:57:39 +0800
  • 9be8de60 fix(midout): formatting midout tools by Megvii Engine Team 2022-05-06 16:11:51 +0800
  • 8182af6e fix(mgb): fix strategy of grad_op and opr_attr by Megvii Engine Team 2022-04-25 11:20:35 +0800
  • 70209667 fix(dnn/test): fix some bug when force_deduce_layout is off by Megvii Engine Team 2022-04-28 14:46:19 +0800
  • 597a1e79 refactor(imperative): add interface to clear algorithm cache by Megvii Engine Team 2022-04-26 21:06:35 +0800
  • e2f5156b refactor(megbrain): save fastrun result to algorithm cache by Megvii Engine Team 2022-04-13 14:40:29 +0800
  • f902ba24 docs(megbrain): add notes for fastrun by Megvii Engine Team 2022-04-21 19:19:32 +0800
  • d968942f perf(cuda): speedup direct large kernel conv by Megvii Engine Team 2022-04-05 16:38:02 +0800
  • b4f9703f fix(overflow): fix the overflow of the long_scalars in network_node and module_stats by Qsingle 2022-05-08 21:05:17 +0800
  • b2cffdde fix(lite): fix lite cpu default not work by Megvii Engine Team 2022-04-26 14:12:54 +0800
  • dbce6526 fix(mge/functional): fix return dtype of comparison function by Megvii Engine Team 2022-04-26 19:31:34 +0800
  • 7dc34769 feat(dnn/cuda): add typecvt uint16 by Megvii Engine Team 2022-04-24 22:18:11 +0800
  • b92866d2 fix(build): fix build depends dirty file issue by Megvii Engine Team 2022-04-24 14:46:30 +0800
  • 9d8983f3 Merge pull request #457 from kagome1007/updatenewlogo by megvii-mge 2022-05-06 14:49:11 +0800
  • e3f591ca fix(mge): update logo by “wenjuan” 2022-05-05 14:48:46 +0800
  • 4b27e861 fix(ops): implement from_op_node for reshape by Megvii Engine Team 2022-04-13 12:18:51 +0800
  • 4fb3d886 perf(misc): use reinterpret_cast to convert valueshape by Megvii Engine Team 2022-04-12 19:48:03 +0800
  • 951ed476 fix(minigraph): supports varnode forwarding by Megvii Engine Team 2022-04-12 19:45:18 +0800
  • 27d4c4b3 refactor(stats): use static inline variable declaration by Megvii Engine Team 2022-04-12 19:26:07 +0800
  • 787a22a9 perf(tensor): implement __new__ in cpp by Megvii Engine Team 2022-04-12 19:51:45 +0800
  • 99df4a79 fix(dtype): dtype scalar set_retain_dtype supports bool by Megvii Engine Team 2022-04-12 19:29:32 +0800
  • 53d48ca3 fix(fastrun): persistent cache add prefix by Megvii Engine Team 2022-04-12 19:30:28 +0800
  • 7005b7d6 fix(mge): fix fastpath check by Megvii Engine Team 2022-04-20 19:43:05 +0800
  • 7bf5b0ee test(imperative): check env values after each pytest by Megvii Engine Team 2022-04-21 17:39:05 +0800
  • 811579b0 fix(imperative): restrict value converts to symbolvar by Megvii Engine Team 2022-04-20 11:21:58 +0800
  • b3f79966 fix(mgb): fix "TRT_ERROR: INVALID_ARGUMENT: Get binding data type failed." by Megvii Engine Team 2022-04-14 16:41:59 +0800
  • 99a85c40 fix(mge): fix advanced indexing grad by Megvii Engine Team 2022-04-18 14:05:55 +0800
  • 409c9881 fix(imperative): add matmul apply_on_varnode by Megvii Engine Team 2022-04-14 10:29:47 +0800
  • d52ba79d fix(lite): support set data by copy on device tensor by Megvii Engine Team 2022-04-13 18:25:08 +0800
  • 275f12c9 fix(mge): fix dimshuffle shape infer by Megvii Engine Team 2022-04-12 13:32:12 +0800
  • e59b6e13 fix(imperative/src): fix empty_tensor bug of convbwd&rng by Megvii Engine Team 2022-04-11 16:20:17 +0800
  • 115c4592 fix(dnn/opencl): fix opencl elemwise tuning issue by Megvii Engine Team 2022-04-01 14:55:05 +0800
  • 275a492b docs(mge/functional): update functional.nn.max_pool2d docstring by Seeker 2022-03-02 15:50:58 +0800
  • a22669c1 docs(mge/functional): update functional.nn.avg_pool2d docstring by Seeker98 2022-03-02 16:24:20 +0800
  • b9cbc101 feat(lite): add pack model by Megvii Engine Team 2022-04-12 17:21:23 +0800
  • 7927e98f perf(mge): speed up PixelShuffle by Megvii Engine Team 2022-03-22 13:05:46 +0800
  • 43e5f41c docs(mgb/docs): add alternative interface for deprecated api by Megvii Engine Team 2022-04-08 16:42:26 +0800
  • ba90a028 feat(mge): opt third_party prepare 2/2 by Megvii Engine Team 2022-04-14 14:19:05 +0800
  • ff9e6121 fix(lite): fix lar LITE_WITH_CUDA and LITE_WITH_OPENCL invalid for bazel by Megvii Engine Team 2022-04-08 19:42:40 +0800
  • 73112558 feat(mge/dnn): support checknonfinite for fp16 by Megvii Engine Team 2022-03-04 18:45:06 +0800
  • f7e10ea8 perf(imperative): improve matmul/batch_matmul by Megvii Engine Team 2022-03-22 18:26:57 +0800
  • 1c2a323e feat(mge): add warning message when mismatched cuda sm is detected by Megvii Engine Team 2022-04-06 17:15:35 +0800
  • 877bda41 perf(mge): improve cross stream memory borrowing by Megvii Engine Team 2022-02-14 14:06:31 +0800
  • ed7fa104 feat(fallback): move direct multi_thread_common helper to fallback by Megvii Engine Team 2022-04-07 14:12:46 +0800
  • 8871ad74 refactor(fallback): opt gi naive reinterpret by Megvii Engine Team 2022-04-07 18:00:29 +0800
  • c2eec47b fix(imperative/src): fix adaptive_pooling bug by Megvii Engine Team 2022-04-09 01:08:50 +0800
  • ffbf8fad feat(fallback): add general intrinsic to elemwise multitype by Megvii Engine Team 2022-03-25 15:13:33 +0800
  • 484e1f11 fix(build): fix riscv64 gcc build with > O0 by Megvii Engine Team 2022-04-06 19:43:42 +0800
  • 14e9ad62 fix(megdnn): emit define-but-not-referenced and extra-;-ignored warning on cuda9.0~cuda9.1 by Megvii Engine Team 2022-04-06 20:56:31 +0800
  • 4c0bff1d refactor(megdnn): refactor TEGRA_X1/X2 macro by Megvii Engine Team 2022-03-26 01:02:05 +0800
  • 758549b9 feat(megengine): support tx2 by Megvii Engine Team 2022-03-19 15:29:43 +0800
  • 84ce94fb docs(imperative): fix docs about vision related function by Megvii Engine Team 2022-03-22 11:03:00 +0800
  • c2435d15 perf(imperative): specialize adaptive pooling by Megvii Engine Team 2022-03-22 19:10:17 +0800
  • 8fcbe825 docs(mge/functional): fix functional.eye docstring by Megvii Engine Team 2022-04-01 17:47:09 +0800
  • 6aa42e6e Merge pull request #439 from Qsingle:fix_python36_pip_url by Megvii Engine Team 2022-04-06 10:33:30 +0800
  • 39d98d45 feat(fallback): add fallback typecvt with general intrinsic by Megvii Engine Team 2022-03-09 18:31:34 +0800
  • d2278f02 perf(imperative): speed up conv_transpose3d by Megvii Engine Team 2022-03-23 15:42:21 +0800
  • 3a5347ed perf(imperative): speed up pooling by Megvii Engine Team 2022-03-23 14:36:48 +0800
  • c0b267ff refactor(cuda-stub): opt cuda-stub log by Megvii Engine Team 2022-03-31 17:51:40 +0800
  • 0bc5a1c8 feat(mge): opt third_party prepare by Megvii Engine Team 2022-03-30 15:58:19 +0800
  • 3f11b421 docs(functional): replace megengine function testcode with doctest format by Megvii Engine Team 2022-04-01 12:09:42 +0800
  • 60163c5d docs(functional): replace elem testcode with doctest format by Megvii Engine Team 2022-02-28 19:59:26 +0800
  • fd5031f8 docs(functional): replace utils testcode with doctest format by Megvii Engine Team 2022-02-28 17:10:19 +0800
  • aa632305 docs(functional): replace loss function testcode with doctest format by Megvii Engine Team 2022-02-28 17:00:17 +0800
  • e8d0f9db docs(example): replace dist.help testcode with doctest format by Megvii Engine Team 2022-02-28 15:14:51 +0800
  • f5f9249a docs(compose): update compose API docstring and example by Megvii Engine Team 2022-02-28 13:57:07 +0800
  • 7d3a6db0 docs(tensor): add more introduction about Tensor by Megvii Engine Team 2022-02-25 17:01:15 +0800
  • 30561d72 fix(windows): fix dll install on windows by Megvii Engine Team 2022-04-01 14:34:51 +0800
  • d9c4ef59 perf(imperative): using simple hash key in heuristic cache by Megvii Engine Team 2022-03-24 19:05:33 +0800
  • 26ea33c6 perf(imperative): improve convbwd performance by Megvii Engine Team 2022-03-31 22:07:18 +0800
  • 3949d425 feat(core): always show MegEngine version and git commit id by Megvii Engine Team 2022-03-29 19:23:12 +0800
  • 5bf31163 (tag: v1.9.0) fix(mge): fix infer output attrs fallible by Megvii Engine Team 2022-03-30 12:43:42 +0800
  • 94960ecf fix(imperative): restrict using convert_inputs in py_apply by Megvii Engine Team 2022-03-30 15:39:46 +0800
  • 09dab387 feat(cuda): support int1 simplewq conv by Megvii Engine Team 2022-03-11 20:02:59 +0800
  • f319d842 fix(scripts): update the pip url for python36 by Qsingle 2022-02-23 11:48:39 +0800
  • 90cedd3c fix(imperative): restrict using convert_inputs in py_apply by Megvii Engine Team 2022-03-30 15:39:46 +0800
  • 7a63f1cd docs(readme): opt cmake build md by Megvii Engine Team 2022-03-29 13:21:44 +0800
  • 2d72de8a fix(mge): fix infer output attrs fallible by Megvii Engine Team 2022-03-30 12:43:42 +0800
  • b6ad4572 feat(cuda): support int1 simplewq conv by Megvii Engine Team 2022-03-11 20:02:59 +0800
  • 331567af fix(opencl/ci): misc opt and fix: by Megvii Engine Team 2022-03-25 17:07:12 +0800
  • ff6a3bb8 fix(fallback): delete the repeat opcaller in fallback and arm_common by Megvii Engine Team 2022-03-09 18:31:34 +0800
  • 547945e8 feat(fallback): support general intrinsic in elemwise in fallback by Megvii Engine Team 2022-03-08 15:47:32 +0800
  • a017bed3 fix(fallback): reman general intrinsic type and add more intrinsic by Megvii Engine Team 2022-03-08 15:45:22 +0800
  • 6554e262 chore(release): bump version by 温娟 2022-03-28 06:17:52 +0000
  • fd6f8e58 feat(mgb/dtype): add dtype qint1 by Megvii Engine Team 2022-03-15 17:13:23 +0800
  • 616352b0 fix(imperative): add dtype promote support for concat by Megvii Engine Team 2022-03-25 13:25:29 +0800
  • 95a30eb6 perf(imperative): speed up stackmanager guard by Megvii Engine Team 2022-03-24 15:14:09 +0800
  • f7a5fe17 feat(mge/config): add a config option for memory forwarding by Megvii Engine Team 2022-03-24 15:27:57 +0800
  • 9ffc2c0a fix(mge): fix host performance loss caused by dtr by Megvii Engine Team 2022-03-24 15:04:21 +0800
  • 69673f14 fix(imperative): remove convert_inputs from concat by Megvii Engine Team 2022-03-24 14:02:59 +0800
  • 4d5faa3f fix(imperative): using DnnOprCaller to avoid early destruction of dnn_opr by Megvii Engine Team 2022-03-22 15:23:57 +0800
  • da620ca1 perf(imperative): specialize batchnorm implementation by Megvii Engine Team 2022-03-21 15:39:19 +0800
  • 5ebc9d50 fix(pylite): fix lite global layout transform and fast run conflict error by Megvii Engine Team 2022-03-07 19:47:06 +0800
  • 49d92d9c feat(lite): feat layout transform interface for lite model by Megvii Engine Team 2022-02-28 18:18:48 +0800
  • 2a900a69 perf(imperative): improve reduce op performance by Megvii Engine Team 2022-03-01 17:31:22 +0800