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

Commit Graph

  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • |\
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • 6ce212d2 refactor(mgb): refactor group conv by Megvii Engine Team 2021-06-30 20:02:52 +0800
  • febd0b17 ci(fix): fail when git user name or email is empty by XindaH 2021-07-27 20:07:46 +0800
  • eb2dd018 build(fp16): fix fp16 build by Megvii Engine Team 2021-07-23 18:28:22 +0800
  • f76a2cc2 feat(mge/opr): add silu and gelu by Megvii Engine Team 2021-07-16 15:44:33 +0800
  • f2ac4c34 docs(distributed.functional.all_reduce_sum): googlestring and examples by Megvii Engine Team 2021-07-23 17:15:50 +0800
  • 186bacfb fix(mge): recover bn freeze fastpath execution by Megvii Engine Team 2021-07-22 14:10:03 +0800
  • 5f558042 fix(imperative/ops): use tblgen to generate FastpathCopy by Megvii Engine Team 2021-07-23 16:23:15 +0800
  • bfc4e7a9 docs(mge): fix amp docstring problems by Megvii Engine Team 2021-07-21 16:47:04 +0800
  • 0b764cf2 docs(mge/functional): add docs for megengine.functional.full_like by Megvii Engine Team 2021-07-19 21:16:44 +0800
  • f1411590 refactor(mge): loose the error bound of fastrun by Megvii Engine Team 2021-07-20 17:06:10 +0800
  • 1f043696 refactor(mge/distributed): using nccl as default in distributed training by Megvii Engine Team 2021-07-19 16:26:52 +0800
  • b17a02d4 feat(mge/distributed): deprecate get_device_count_by_fork by Megvii Engine Team 2021-07-19 14:54:45 +0800
  • f8b0f2cb build(dnn/cutlass): fix build for cutlass by Megvii Engine Team 2021-07-19 14:25:58 +0800
  • 0fb4e9a9 fix(ci): git set user and email by konghuanjun 2021-06-24 14:20:40 +0800
  • 6af4a32e feat(mge/third_party): update MegRay version by huangxinda 2021-07-19 20:53:10 +0800
  • 093f7ae7 feat(mge/third_party): update cutlass version by huangxinda 2021-07-19 17:12:41 +0800
  • c2daea3c chore(release): bump version by Megvii Engine Team 2021-07-18 16:39:40 +0800
  • 207a3463 chore(mge): run get_device_count("gpu") in subprocess by Megvii Engine Team 2021-01-19 18:27:50 +0800
  • 869a0327 perf(mgb): disable FoldingConvBiasDimshufflePass in cuda10 for performance by Megvii Engine Team 2021-07-13 13:13:03 +0800
  • 0baf6b0d Merge pull request #175 from tpoisonooo:fix-spell-error by liuke 2021-07-23 18:14:07 +0800
  • 239916a9 fix(mgb/gopt): fix testcase for enable nchw64 pass by Megvii Engine Team 2021-06-16 16:32:36 +0800
  • 2ab5c53f feat(mgb/gopt): support nhwc conv in tensor reformat pass by Megvii Engine Team 2021-06-09 14:54:29 +0800
  • 009c90a2 feat(mgb/gopt): modify padding policy for 4bit conv bias oprs by Megvii Engine Team 2021-06-08 14:30:19 +0800
  • 4eda3388 feat(dnn/cuda): generate cutlass kimpls using cmake and bazel by Megvii Engine Team 2021-06-21 17:53:19 +0800
  • 8d248a6a fix(dnn/cuda): fix testcase for fallback nchw qs8 conv by Megvii Engine Team 2021-06-23 15:18:43 +0800
  • 894a2407 feat(dnn/cuda): add relayout format kernel for nchw <-> nhwc by Megvii Engine Team 2021-06-04 14:43:59 +0800
  • 43c59204 refactor(dnn/cuda): refactor relayout format kernels by Megvii Engine Team 2021-05-31 19:01:42 +0800
  • f41a8086 feat(dnn/cuda): add nhwc int4 conv support by Megvii Engine Team 2021-06-11 14:48:54 +0800
  • 5a14a892 refactor(dnn/cuda): refactor cutlass kernel generator for gemm and gemv by Megvii Engine Team 2021-05-31 11:41:30 +0800
  • b33217d8 refactor(dnn/cuda): refactor cutlass kernel generator for deconv operation by Megvii Engine Team 2021-05-27 19:32:24 +0800
  • 4abf7bd3 refactor(dnn/cuda): refactor kernel generator for cutlass convolution kernels by Megvii Engine Team 2021-05-27 13:35:25 +0800
  • b4687ce8 feat(dnn/cuda): add convolution with i8 input and u4 output by Megvii Engine Team 2021-05-24 19:03:53 +0800
  • 00083d13 fix(dnn/cuda): fix recursive algo search for fallback_nchw_qs8 by Megvii Engine Team 2021-05-20 14:41:06 +0800
  • bba04f02 feat(mgb/gopt): add fusion support for conv, astype(s4) and reformat by Megvii Engine Team 2021-05-18 20:46:10 +0800
  • 66f70578 feat(dnn/cuda): add convolution with i8 input and i4 output by Megvii Engine Team 2021-05-18 19:08:22 +0800
  • 6d686ff2 feat(gopt/inference): allow Float32 output dtype in EnableNCHW64Pass by Megvii Engine Team 2021-06-16 15:56:03 +0800
  • 7d3df995 feat(gopt/inference): allow Float32 output dtype in EnableNCHW4Pass by Megvii Engine Team 2021-06-11 18:38:56 +0800
  • 633016a9 fix(dnn/cuda): fix AlgoFallbackNCHWQS8 to support Float32 dst by Megvii Engine Team 2021-06-10 18:30:41 +0800
  • e6caa9ff feat(opr): add bn backward for inference mode by Megvii Engine Team 2021-06-01 12:36:09 +0800
  • c90fa087 test(mge): delete test_external.py by Xinda Huang 2021-07-17 15:46:22 +0800
  • b2944559 fix(imperative/module): remove ``__getattribute__`` method in module by Megvii Engine Team 2021-07-03 12:25:02 +0800
  • 77ead937 fix(src/serialization): fix compatibility error of oss model by Megvii Engine Team 2021-07-13 14:32:05 +0800
  • 070c8117 fix(imperative): remove convert_inputs by Megvii Engine Team 2021-07-07 10:15:13 +0800
  • f40df602 docs(mge): refactor docs to remove warnings by Megvii Engine Team 2021-07-09 14:34:20 +0800
  • 1040b778 fix(mge/functional): fix F.topk(kth_only=True) by Megvii Engine Team 2021-06-28 19:42:16 +0800
  • 551cc701 docs(distributed.functional): add return type for all_reduce_max (jira #MGE-2706) by Megvii Engine Team 2021-07-13 18:51:53 +0800
  • 72ff7aec feat(docs): add docs for megengine.functional.ones_like(jira #MGE-2702) by Megvii Engine Team 2021-07-13 23:56:29 +0800
  • 7c9569e4 fix(mge/random): fix random seed by Megvii Engine Team 2021-07-09 14:33:09 +0800
  • 07de1571 fix(mgb): remove static mem record from tee by Megvii Engine Team 2021-07-06 15:43:35 +0800
  • d7b6bfd5 test(mge/fakequant): use fixed input for lsq test to temperarily avoid precision error by Megvii Engine Team 2021-07-14 11:23:16 +0800
  • 5cef74a7 feat(mge/amp): add GradScaler support by Megvii Engine Team 2021-06-29 18:17:44 +0800
  • 1bf18252 feat(mge/amp): add mix precision autocast support by Megvii Engine Team 2021-05-11 18:55:40 +0800
  • f12355f7 fix(imperative/grad): fix hardcode dtype in subtensor_grad_rule by Megvii Engine Team 2021-05-18 10:46:48 +0800
  • 4e4497b9 refactor(mgb/dnn): x86 pooling rebase algochooser by Megvii Engine Team 2021-07-12 17:01:04 +0800
  • a33c3b73 refactor(mgb/dnn): arm pooling rebase algochooser by Megvii Engine Team 2021-07-11 22:52:32 +0800
  • 8dea6b3c build(dnn): compat for more windows env by Megvii Engine Team 2021-07-08 14:30:14 +0800
  • 56b94d89 feat(dtr): add sqrt sampling by Megvii Engine Team 2021-07-13 03:01:41 +0800
  • 8a73193c feat(dtr): remove eviction threshold by Megvii Engine Team 2021-07-12 18:16:53 +0800
  • 69d1fd0f refactor(opdef): split apply_on_physical_tensor into infer_output_mem_desc and execute by Megvii Engine Team 2021-07-12 17:38:36 +0800
  • 75eb04c5 feat(mge/experimental): add WeightScaler support by Megvii Engine Team 2021-07-01 14:16:07 +0800
  • dedecf69 fix(imperative/utils): fix logical error of replace var by Megvii Engine Team 2021-06-23 18:12:47 +0800
  • ea70d99b fix(mge/convbias): make fallback convbias support nhwcd4 layout by Megvii Engine Team 2021-06-29 13:46:17 +0800
  • 497ef6c3 fix(mge/dist): fix gl oom error by Megvii Engine Team 2021-07-13 12:51:02 +0800
  • 43098fb8 feat(mge): add SlidingWindowTranspose opr by Megvii Engine Team 2021-06-29 14:19:59 +0800
  • df79334c feat(mge/distributed): add user_pop function to save device memory by Megvii Engine Team 2021-07-07 20:35:06 +0800
  • 1eaf32cd fix(mgb): fix typo in message by Megvii Engine Team 2021-07-05 10:32:14 +0800
  • 7225b0f0 fix(mge/utils): use static infer manager to get value of network.varnode by Megvii Engine Team 2021-06-10 17:29:19 +0800
  • ffe2bb2e fix(mge): fix some errors caused by unknown shape when using symbolic trace or building graph by Megvii Engine Team 2021-06-10 17:24:43 +0800
  • 2d42455f fix(mge/utils): fix toposort to get definition order by Megvii Engine Team 2021-06-01 19:15:43 +0800
  • 0c97b2a3 fix(module): remove assert during forward by Megvii Engine Team 2021-07-08 19:56:38 +0800
  • 42711308 fix(module/normalization): fix bug of LayerNorm and support input of any shape by Megvii Engine Team 2021-05-14 20:40:22 +0800
  • a95f6d4f perf(trace): add fastpath for const value assert by Megvii Engine Team 2021-07-11 01:45:34 +0800
  • 2cd98232 fix(mgb/tensorrt): fix trt runtime, padding channel to a multiple of 4 when using kCHW4 IOFormat by Megvii Engine Team 2021-07-07 18:56:40 +0800
  • b078dda9 feat(mge/random): add some random op and remove random/distrbution.py by Megvii Engine Team 2021-06-07 16:44:20 +0800
  • 83e4c9d7 fix(opencl): open opencl topk test when opencl beyond 2.0 by Megvii Engine Team 2021-07-05 14:12:39 +0800
  • f30c0e06 feat(mgb/opr): add lsq opr by Megvii Engine Team 2021-06-28 13:14:08 +0800
  • eb66681f fix(mge/random): fix delete_rng_handel by Megvii Engine Team 2021-07-08 20:30:21 +0800
  • d6db4fea fix(mge/module): set no_cache=true when loading state dict by Megvii Engine Team 2021-07-01 13:51:11 +0800
  • fea1bba2 fix(mge/tools): fix module status error by Megvii Engine Team 2021-07-08 15:32:38 +0800
  • 6cd01d5a feat(imperative/functional): let elemwise support empty IO & add some tests by Megvii Engine Team 2021-06-25 16:36:55 +0800
  • dea52781 feat(mgb/opr): let PowC & TypeCvt support empty IO by Megvii Engine Team 2021-06-25 16:34:58 +0800
  • 2f68aeb9 feat(imperative/jit): let trace support empty IO by Megvii Engine Team 2021-06-25 16:27:56 +0800
  • 809d5056 feat(mge/distributed): enable pt shm allreduce by Megvii Engine Team 2021-03-10 16:50:44 +0800
  • 02455941 test(autograd): test jvp emulated by 2nd grad by Megvii Engine Team 2021-07-07 14:16:09 +0800
  • 8480302d fix(autograd): make higher order grad experimental by Megvii Engine Team 2021-07-06 13:16:43 +0800
  • 72531f2b test(autograd): add more tests for higher order grad by Megvii Engine Team 2021-06-18 14:40:24 +0800
  • 522e556b feat(autodiff): support higher order grad by Megvii Engine Team 2021-05-24 16:56:46 +0800
  • 5198b783 fix(mge/functional): fix expand_dims for scalar by Megvii Engine Team 2021-07-02 15:28:45 +0800
  • 88898e63 fix(mgb): replace if_constexpr with runtime function to avoid potential bug by Megvii Engine Team 2021-07-05 15:54:02 +0800
  • 25932352 refactor(mgb/dnn): rocm pooling rebase algochooser by Megvii Engine Team 2021-06-15 13:53:39 +0800
  • 1cfdbc56 feat(dnn): add deterministic max pooling by Megvii Engine Team 2021-05-19 17:49:45 +0800
  • 20ab82d0 fix(tee): fix tee crash by Megvii Engine Team 2021-07-05 19:00:11 +0800
  • 933dd9a4 feat(mge/distributed): add cuda env check before forked thread by Megvii Engine Team 2021-06-29 16:37:01 +0800
  • 2a541961 fix(tee): fix tee link by Megvii Engine Team 2021-07-01 23:09:30 +0800
  • a5060a2b feat(mgb/opr): add check_has_inf kernel and opr by Megvii Engine Team 2021-06-23 18:38:22 +0800
  • 3597a6db feat(dnn/arm): nchw_nchw44 conv support 1x1s1 by Megvii Engine Team 2021-06-23 21:18:09 +0800
  • c64b1c94 feat(imperative/functional): add roll in functional by Megvii Engine Team 2021-05-25 14:12:27 +0800
  • d915c5a3 refactor(mgb): make convolution3D handle noncontiguous tensors by Megvii Engine Team 2021-06-24 14:18:02 +0800
  • d04cd67f refactor(mgb): make conv-backward-filter handle noncontiguous tensors by Megvii Engine Team 2021-06-11 19:17:20 +0800
  • 44376f70 refactor(mgb): make conv-backward-data handle noncontiguous tensors by Megvii Engine Team 2021-06-11 18:24:29 +0800