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

Commit Graph

  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • | *
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • |/
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • |\
  • | *
  • |/
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • |\
  • | | *
  • | | *
  • | | *
  • | | *
  • | | *
  • | | *
  • | * |
  • |/ /
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • * |
  • 8546c15d feat(gi): make elemwise apply gi class type by Megvii Engine Team 2022-06-14 18:14:48 +0800
  • 74fb63db feat(gi): make matrix_mul apply gi class type by Megvii Engine Team 2022-06-14 18:13:40 +0800
  • 45b26400 feat(gi): make resize apply gi class type by Megvii Engine Team 2022-06-14 18:12:21 +0800
  • 7d7cc3c8 feat(gi/riscv): add gi support with risc-v by Megvii Engine Team 2022-06-14 18:09:59 +0800
  • a32b7277 fix(build): upgrade bazel riscv toolchains by Megvii Engine Team 2022-06-14 18:09:08 +0800
  • 588a645b feat(dnn/opencl): add binary interface for opencl algo and kernel cache by Megvii Engine Team 2022-06-10 18:42:39 +0800
  • 1192a9a6 fix(imperative): fix adaptive pool2d by Megvii Engine Team 2022-06-14 20:52:27 +0800
  • 7b4b94fd fix(imperative): fix the segmentfault when reduce backward by Megvii Engine Team 2022-06-13 16:25:17 +0800
  • 24c5c19b fix(imperative): make functional ops support negative axis by Megvii Engine Team 2022-04-22 16:26:30 +0800
  • c76e80bc fix(mge): fix mem leak of numpy method of DeviceTensorND by Megvii Engine Team 2022-06-14 11:09:52 +0800
  • f96429c0 feat(imperative): support empty tensor in roi_align by Megvii Engine Team 2022-05-05 13:36:40 +0800
  • 2f829aaa test(imperative): speed up integration test by Megvii Engine Team 2022-05-13 16:09:41 +0800
  • a1ca390d fix(lite): fix const shape error for lar fitting mode by Megvii Engine Team 2022-06-08 15:46:41 +0800
  • 2001c494 feat(lite): add input shape parse for load and run by Megvii Engine Team 2022-05-26 20:02:32 +0800
  • 64a8aaaf fix(build): remove ununsed functions when cuda disabled by Megvii Engine Team 2022-06-10 16:53:15 +0800
  • 8f17b84a fix(dnn): fix dnn run cd4 on cpu by Megvii Engine Team 2022-06-07 14:22:56 +0800
  • 81065cf0 build(mgb/cutlass): merge partial headers by Megvii Engine Team 2022-04-30 11:55:36 +0800
  • d610c987 feat(lite): add disable configure by model info interface by Megvii Engine Team 2022-06-07 17:08:48 +0800
  • 07bdb3bf feat(imperative): add swapaxes by Megvii Engine Team 2022-06-06 17:02:59 +0800
  • a0862865 feat(mge/third_party): update cutlass version by huangxinda 2022-06-13 11:56:51 +0800
  • 972a8d54 feat(ci): add model compatibility check in ci by Megvii Engine Team 2022-05-20 20:04:43 +0800
  • e6943017 fix(ops/jit): skip lookup include path when nvcc executable not found by Megvii Engine Team 2022-06-08 19:11:42 +0800
  • 02c1a0c3 fix(imperative/format): warn once when parameter without format been attached by Megvii Engine Team 2022-06-07 15:48:12 +0800
  • 6ef1e12c fix flops count bug for ConvTranspose2d by fanhqme2 2022-06-09 20:07:48 +0800
  • c2deef1a feat(mge): aad atlas710 support by Megvii Engine Team 2022-05-31 20:50:55 +0800
  • df5ebd3d fix(imperative/ops): fix the vmemory problem in 1.9 by Megvii Engine Team 2022-06-08 16:13:19 +0800
  • 4e66e0eb feat(megdnn/softmax): add softmax operator in OpenCL by Megvii Engine Team 2022-05-06 16:53:34 +0800
  • b36b5bd8 refactor(mgb): check input when profiling by Megvii Engine Team 2022-06-01 11:10:43 +0800
  • 6c9b3a58 refactor(dnn): remove algorithm cache queries by Megvii Engine Team 2022-05-10 13:34:39 +0800
  • 8563f514 fix(imperative): fix buildin reduce keepdim by Megvii Engine Team 2022-06-06 11:51:23 +0800
  • 96d90be1 feat(dnn): fallback support int4 relayout by Megvii Engine Team 2022-06-01 19:40:01 +0800
  • eef0308b feat(serialization): add no_change_graph and version param whem dump model by Megvii Engine Team 2022-06-01 14:43:39 +0800
  • 4ab5f970 fix(build): fix ci error by Megvii Engine Team 2022-05-12 18:57:04 +0800
  • b9a69323 feat(imperative): channel default model format to fbs v2 by Megvii Engine Team 2022-05-12 15:42:09 +0800
  • 283a2f77 feat(serialization): support test for new serialization format by Megvii Engine Team 2022-05-11 18:03:28 +0800
  • 50faabf6 feat(serialization): support the registry for new serialization format by Megvii Engine Team 2022-05-11 18:02:52 +0800
  • a694fb33 feat(serialization): implement the new serialization format by Megvii Engine Team 2022-05-11 17:57:46 +0800
  • ca4a5da0 feat(serialization): add new serialization format define by Megvii Engine Team 2022-05-11 17:54:48 +0800
  • ba32360a feat(lite): add set opencl buffer kernel cache lite api by Megvii Engine Team 2022-06-07 13:36:13 +0800
  • 711b5bf5 fix(dnn/arm_common): fix some load beyond memory by Megvii Engine Team 2022-05-17 16:49:44 +0800
  • 3ebb8db0 feat(third_party/cutlass): update to version 2.8 by Megvii Engine Team 2022-04-27 17:58:37 +0800
  • da91e650 refactor(ops/layer_norm): speed up the host speed of layer_norm by Megvii Engine Team 2022-05-24 21:33:48 +0800
  • 67cfce9f fix(imperative/amp): add is_scalar check in elemwise and concat by Megvii Engine Team 2022-05-31 14:37:55 +0800
  • d313f926 fix(imperative/amp): fix format transformation for symbol trans by Megvii Engine Team 2022-05-28 16:56:43 +0800
  • 261a5bce feat(imperative/amp): add dimshuffle in set_format for nhwc by Megvii Engine Team 2022-05-26 18:00:29 +0800
  • c9e56f49 feat(imperative/amp): add dimshuffle before creating nhwc tensor by Megvii Engine Team 2022-05-26 15:36:27 +0800
  • d57a0712 feat(imperative/amp): add fallback for op not supported for nhwc tensor by Megvii Engine Team 2022-05-19 22:30:09 +0800
  • 38a9aa9f feat(imperative/amp): add auto dimshuffle for elemwise and concat by Megvii Engine Team 2022-05-16 15:14:59 +0800
  • cd263765 style(imperative/amp): reformat code by Megvii Engine Team 2022-05-06 14:58:01 +0800
  • 3892aa0b fix(imperative/amp): fix bn params for nhwc amp by Megvii Engine Team 2022-04-29 14:48:52 +0800
  • 6f0b5820 chore(imperative/amp): adapt dev by Megvii Engine Team 2022-04-22 14:57:21 +0800
  • ee984e86 fix(imperative/amp): fix distributed backward callback for nhwc amp by Megvii Engine Team 2022-04-11 18:44:25 +0800
  • 15c6da62 feat(imperative/amp): add nhwc support for adaptive pooling by Megvii Engine Team 2022-04-13 18:09:58 +0800
  • c28a875f fix(imperative/amp): adapt new transformation by Megvii Engine Team 2022-03-29 17:11:31 +0800
  • fd41302c feat(imperative/amp): add set_format by Megvii Engine Team 2022-03-25 11:01:21 +0800
  • fc633ce4 fix(imperative/amp): fix custom grad in Subgraph by Megvii Engine Team 2022-03-21 19:55:16 +0800
  • 673b295d feat(imperative/amp): remove conv_format and bn param_dim configs by Megvii Engine Team 2022-03-11 12:44:23 +0800
  • 7e9aa742 feat(imperative/amp): enable auto_convert_format by default by Megvii Engine Team 2022-03-08 19:05:55 +0800
  • fc0f4546 fix(dnn/check_non_finite): adjust some details of CheckNonFinite by Megvii Engine Team 2022-02-24 14:59:56 +0800
  • 3bd40887 feat(mgb/opr): add NHWC support for AdaptivePooling by Megvii Engine Team 2022-02-11 14:37:14 +0800
  • e393d1cf feat(mge/amp): add convert_format module for NHWC training by Megvii Engine Team 2022-01-25 18:00:41 +0800
  • 533fb5bf feat(imperative): support formatted tensor and add special op rules by Megvii Engine Team 2022-01-07 16:48:40 +0800
  • 4aa79c45 perf(mge): override grad of matmul by Megvii Engine Team 2022-05-12 17:14:16 +0800
  • 98b5ee78 feat(mge/dnn): add lamb optimizer by Megvii Engine Team 2022-04-07 18:37:31 +0800
  • a926878c feat(imperative): remove symbolvar of imperative by Megvii Engine Team 2022-05-15 20:52:59 +0800
  • 14813d13 fix(whl): fix whl broken: patchelf on big (> 4G) file will make elf section broken, as a workaround, do strip firstly, then do patchelf. by Megvii Engine Team 2022-05-26 01:14:38 +0800
  • 9e0583e1 feat(dnn/arm_common): add arm_common chanwise dot 11x11 by Megvii Engine Team 2022-05-24 14:14:25 +0800
  • 115bcbce feat(lite): add fitting mode for load and run by Megvii Engine Team 2022-05-24 17:13:20 +0800
  • 02bfb8f8 feat(lite): add and fix some feature for load and run fitting mode by Megvii Engine Team 2022-05-17 14:03:15 +0800
  • c62ddba2 feat(dnn/opencl): optimize heuristic rule by Megvii Engine Team 2022-04-01 14:55:05 +0800
  • 6e839407 Merge pull request #461 from tpoisonooo/patch-1 by Xinran Xu 2022-06-06 20:50:10 +0800
  • 91a45d7c docs(README.md): add link by tpoisonooo 2022-05-25 11:37:08 +0800
  • d404ed18 feat(ci): update cpuinfo by huangxinda 2022-05-24 14:40:25 +0800
  • c2500cdb chore(license): apply change caused by bot forward rebase by Megvii Engine Team 2022-05-24 10:50:33 +0800
  • 5f0e7ffb feat(fallback): add FB_GI_F32_4x12 benchmark by Megvii Engine Team 2022-05-19 18:46:23 +0800
  • f249d387 feat(fallback): imp gi matmul FB_GI_F32_4x12 algo by Megvii Engine Team 2022-05-17 16:36:56 +0800
  • 03f78547 feat(dnn/arm_common): add 9x9s1s2 dot chanwise kernel by Megvii Engine Team 2022-05-10 16:57:13 +0800
  • 80e1f38b fix(gtest): fix ci error report stack-use-after-scope how to reproduce the problem: 1: build with asan(revert this MR) 2: then taskset process to one cpu: taskset 01 ./megbrain_test --gtest_filter=TestAsyncQueue.SynchronizerWaiterStarving by Megvii Engine Team 2022-05-23 19:08:12 +0800
  • c2e9860f chore(license): remove all license in file header by Megvii Engine Team 2022-02-25 18:44:59 +0800
  • 38b49272 fix(opr): fix no update ptr in reduce operator when input change by Megvii Engine Team 2022-05-16 13:36:45 +0800
  • 4cce2480 fix(dnn/opencl): fix some bug for dnn opencl conv bias and relayout format by Megvii Engine Team 2022-05-19 15:53:46 +0800
  • ca0e616f refactor(lite): refactor load_and_run profiling message by Megvii Engine Team 2022-05-13 14:25:51 +0800
  • 1783b897 feat(profiler): integrate cupti backend by Megvii Engine Team 2022-04-12 19:23:57 +0800
  • e98049d7 feat(fallback): move arm_common resize f32 algo to fallback gi by Megvii Engine Team 2022-05-16 14:46:08 +0800
  • 5b69af20 Merge pull request #460 from kagome1007/updatereadme by megvii-mge 2022-05-20 11:12:25 +0800
  • 0bed6c0f (tag: v1.9.1, release-1.9) chore(release): bump version by 温娟 2022-05-19 05:44:33 +0000
  • 9488cd1c fix(lite): fix lite cpu default not work by Megvii Engine Team 2022-04-26 14:12:54 +0800
  • 518c7f37 fix(imperative/src): fix empty_tensor bug of rng by Megvii Engine Team 2022-04-11 16:20:17 +0800
  • cca38c4e fix(mge): fix fastpath check by Megvii Engine Team 2022-04-20 19:43:05 +0800
  • b9e850a4 test(imperative): check env values after each pytest by Megvii Engine Team 2022-04-21 17:39:05 +0800
  • d984be59 fix(imperative): restrict value converts to symbolvar by Megvii Engine Team 2022-04-20 11:21:58 +0800
  • 824af20b fix(mge): update readme by “wenjuan” 2022-05-18 16:25:49 +0800
  • 6814cf1c fix(lite): fix lite test error by Megvii Engine Team 2022-05-12 17:56:27 +0800
  • 7c8f1847 fix(dnn/x86): fix x86 pooling exec by Megvii Engine Team 2022-05-12 22:55:30 +0800
  • 91aaafd5 feat(fallback): move arm_common pooling f32 algo to fallback gi by Megvii Engine Team 2022-05-11 14:37:41 +0800
  • bde2efa3 feat(lite/load_and_run): support put and get model redis cache by Megvii Engine Team 2022-04-15 19:54:07 +0800
  • 48526abb fix(mgb): fix concat cd4 tensor check size invalid by Megvii Engine Team 2022-04-28 19:54:10 +0800
  • c87d998e feat(mgb): add interface to support opencl IO zero copy when inference by Megvii Engine Team 2022-03-09 15:46:14 +0800
  • af6cdb20 feat(fallback): fix ci by Megvii Engine Team 2022-05-07 18:50:29 +0800
  • e4cc85e5 feat(fallback): move arm_common f32 convbias to fallback gi by Megvii Engine Team 2022-05-06 15:06:00 +0800