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

Commit Graph

  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • *
  • 5dbf218d feat(dnn/x86): add sse 8816 matmul by Megvii Engine Team 2020-06-12 18:28:37 +0800
  • 25b6a131 feat(dnn/x86): add x86 avx2 8x8x16 matmul by Megvii Engine Team 2020-06-12 18:28:37 +0800
  • 273f891b fix(mgb/gopt): fix run-time winograd-transform and nchwxx error by Megvii Engine Team 2020-06-10 17:28:00 +0800
  • 02abc36e fix(mbg/arm_common): fix nchw44-dot misc issue by Megvii Engine Team 2020-06-11 19:59:19 +0800
  • 9ed3882a fix(opr/dnn): fix winograd fast run mismatch by Megvii Engine Team 2020-06-05 15:39:09 +0800
  • 18be23f3 fix(mbg/gopt): fix nchwxx gopt with no fuse conv_bias and winograd fast-run by Megvii Engine Team 2020-06-01 20:24:11 +0800
  • 38f7cbd9 fix(mge/module): fix redundant recursion in `train()` by Megvii Engine Team 2020-06-12 12:53:52 +0800
  • 5c232352 test(mge/quantization): add `quantize_disabled` related test by Megvii Engine Team 2020-06-12 12:00:24 +0800
  • ab913025 feat(mge/quantization): add `quantize_disabled` attribute in Module by Megvii Engine Team 2020-06-11 17:11:39 +0800
  • f4ead788 feat(mgb): static allocation with given padding by Megvii Engine Team 2020-05-26 20:28:29 +0800
  • 575a6dca ci(docker): change object files permission by Megvii Engine Team 2020-06-11 19:03:51 +0800
  • 08ac685e feat(mge/functional): add logsumexp by Megvii Engine Team 2020-06-10 03:42:37 +0800
  • 65ec4f7c fix(ci): fix test timeout by Megvii Engine Team 2020-06-10 10:20:41 +0800
  • ea6bfe6c fix(dnn/cuda-stub): simplify and use proper search paths by Megvii Engine Team 2020-05-28 18:20:05 +0800
  • 01092feb feat(mgb): add PackAllReducePass by Megvii Engine Team 2020-06-01 13:43:26 +0800
  • c7e6c658 refactor(mge/distribute): use is_root (and rank) in stead of rank and root at collective comm by Megvii Engine Team 2020-05-29 19:39:41 +0800
  • ff308e3b feat(mgb/comp_node): generate uid for cuda comp node by Megvii Engine Team 2020-06-08 11:12:49 +0800
  • 32c86211 fix(dnn/cuda): enable cuda algos for nchw quantized by Megvii Engine Team 2020-06-04 11:15:21 +0800
  • b8d8886e refactor(mge/tensor): combine Dict and TensorDict by Megvii Engine Team 2020-06-08 17:53:53 +0800
  • 7751a067 docs(mge/tensor): add advanced index related docs by Megvii Engine Team 2020-06-08 17:51:55 +0800
  • 7b0dbe6a fix(dnn/arm): fix stride 1 support for int8 nchw_nchw44 by Megvii Engine Team 2020-06-03 22:38:53 +0800
  • 198f3eb5 fix(dnn/arm): fix fp32 nchw44 direct workspace bug by Megvii Engine Team 2020-06-04 19:29:40 +0800
  • 49fdddef fix(gopt): fix reorder arith chain pass by Megvii Engine Team 2020-06-04 17:52:03 +0800
  • 6742a58b fix(quant): observer do not use cond_take by Megvii Engine Team 2020-06-09 18:01:24 +0800
  • 9e876203 feat(dnn): add int8 direct conv dot nchw44 by Megvii Engine Team 2020-05-21 23:17:44 +0800
  • 09ceaaae fix(dnn/arm): stride1 support for nchw_nchw44 fp32 conv by Megvii Engine Team 2020-06-03 12:47:13 +0800
  • 50db9b84 fix(gopt): fix paramfuse if the endpoint is const by Megvii Engine Team 2020-05-27 12:42:26 +0800
  • 35bc0e1f fix(mge/function): do not deeply copy saved tensor in Function by Megvii Engine Team 2020-06-01 18:22:43 +0800
  • 47377c7b fix(core): fix memory defragmenter by Megvii Engine Team 2020-06-05 12:13:39 +0800
  • f56f187f fix(mbg/gopt): fix nchw44-dot channel wise trans to nchw44 by Megvii Engine Team 2020-06-04 20:23:04 +0800
  • af29fcb2 feat(mgb/plugin): add param json func for indexing oprs by Megvii Engine Team 2020-06-01 17:07:11 +0800
  • 62753c4d fix(mge/sdk): fix comp_node bug in dump_with_testcast_mge by Megvii Engine Team 2020-06-03 16:55:34 +0800
  • f1c86606 fix(dnn/cuda): fix FuseConvBiasWithZ pass for HSwish activation by Megvii Engine Team 2020-06-04 16:06:29 +0800
  • adfa4688 fix(mge/functional): fix scatter doctest failed for GPU platform issue by Megvii Engine Team 2020-06-04 13:03:21 +0800
  • 4f8e6080 feat(dnn): fix Werror by adding macro by Megvii Engine Team 2020-06-04 15:21:00 +0800
  • d7bb62cf refactor(mgb): move mm_handler from python module into opr-mm by Megvii Engine Team 2020-06-02 15:52:33 +0800
  • 84068a6b fix(mge/data): fix typos in voc and objects365 by Megvii Engine Team 2020-06-02 15:03:36 +0800
  • 3966bb08 feat(dnn/test): split cpu.convolution by Megvii Engine Team 2020-06-01 18:53:28 +0800
  • 8f87a3e9 feat(dnn/arm_common): add int8 nchw44 winograd f23_4x4 f23_8x8 compute float32/int16 output int8 by Megvii Engine Team 2020-05-06 19:48:00 +0800
  • 8ffed043 fix(dnn/x86): fix matrix_mul quantized performance on vnni by Megvii Engine Team 2020-06-02 13:59:16 +0800
  • 1d860f4d fix(dnn/x86): fix dnnl int8 algo on vnni by Megvii Engine Team 2020-06-01 16:56:15 +0800
  • 871e6a51 feat(dnn/x86): opt x86 quantized heuristic by Megvii Engine Team 2020-05-28 19:16:55 +0800
  • 6c29548d fix(dnn/arm): fix nchw_nchw44 dot stride1 support by Megvii Engine Team 2020-06-02 16:07:50 +0800
  • 02cbb13b fix(dnn/arm): fix nchw44 fp32 direct algo oh block and unused stride2 algo by Megvii Engine Team 2020-06-02 16:07:50 +0800
  • d2f5874a fix(mge/module): fix non-str key error of dict in module by Megvii Engine Team 2020-05-26 18:19:46 +0800
  • 30b3d3aa fix(dnn/gopt): add convolution nchw44-dot format gopt by Megvii Engine Team 2020-05-28 15:58:45 +0800
  • 48d1ac14 fix(dnn/arm): fix consistence between create_conv1x1_strategy and can_create_conv1x1_strategy by Megvii Engine Team 2020-05-29 21:51:16 +0800
  • 6d0d5e5a fix(mge/redis_server): fix bug for external connection by Megvii Engine Team 2020-06-01 23:21:35 +0800
  • 15d1bd4d fix(mge/quantization): use list rather than generator to modify submodules by Megvii Engine Team 2020-06-01 15:58:54 +0800
  • 7b2c5a73 fix(mge/quantization): fix histogram observer load and store issue by Megvii Engine Team 2020-06-01 11:32:41 +0800
  • e6820b91 feat(mge/module): add conv and conv_relu quantization module by Megvii Engine Team 2020-05-28 14:07:24 +0800
  • a1f8ecc7 fix(dnn/naive): add convolution nchw44-dot format by Megvii Engine Team 2020-05-27 17:26:57 +0800
  • 73d84162 feat(dnn/aarch64): add matmul with dotprod for mk4 by Megvii Engine Team 2020-05-21 14:32:40 +0800
  • c1397792 feat(dnn): add winograd-fp32-nchw44 support by Megvii Engine Team 2020-05-09 20:07:05 +0800
  • 40412e26 feat(mge/module): add sync bn by Megvii Engine Team 2020-05-29 22:00:42 +0800
  • 3c32ad6d feat(dnn/x86): imp avx2 int8 stride2 chanwise conv by Megvii Engine Team 2020-05-26 19:01:08 +0800
  • 89374521 fix(dnn/arm_common): add nchw44 float channel wise s1/s2 by Megvii Engine Team 2020-05-20 17:26:18 +0800
  • 9f997ac5 fix(dnn/x86): enable i8i8i16 gemv used in conv by Megvii Engine Team 2020-05-26 12:18:38 +0800
  • 36e3bb6e feat(mgb/dnn): add armv7 mk4_dot matmul by Megvii Engine Team 2020-05-24 19:32:12 +0800
  • 580a2753 feat(dnn/arm): add nchw44 fp32 direct stride 1 by Megvii Engine Team 2020-05-20 20:47:21 +0800
  • caf1fac2 refactor(mge/quantization): split `QATModule` and refactor convert api by Megvii Engine Team 2020-05-25 19:17:05 +0800
  • ad3c9315 feat(dnn/arm): add arm nchw44 fp32 pooling by Megvii Engine Team 2020-05-21 20:54:47 +0800
  • d19a6379 fix(python_module): hold GIL during _get_serialized_dtype by Megvii Engine Team 2020-05-25 16:03:27 +0800
  • ff341cb1 fix(mge/quantization): modify observer api by Megvii Engine Team 2020-05-25 18:02:33 +0800
  • 27ef788f feat(dnn/armv7): add armv7 mk4 matmul by Megvii Engine Team 2020-05-17 16:35:19 +0800
  • efb60be2 fix(mge/distributed): fix get_free_ports return value by Megvii Engine Team 2020-05-26 14:10:58 +0800
  • 1d7fceca feat(mge/serialization): add map location by Megvii Engine Team 2020-05-01 20:32:42 +0800
  • 9320bf92 feat(mgb/dnn): add matmul mk4 dot naive test by Megvii Engine Team 2020-05-22 14:16:48 +0800
  • a6bc250d feat(dnn/common): add matmul impl for naive with matrix format mk4_dot by Megvii Engine Team 2020-05-22 13:15:45 +0800
  • bb872965 feat(mge/module): add local conv2d module by Megvii Engine Team 2020-05-19 18:26:08 +0800
  • eee3e559 feat(mge/distributed): add multiprocess launcher by Megvii Engine Team 2020-05-08 18:21:18 +0800
  • cde055e8 fix(mgb/opr-mm): fix user_data thread safety in CollectiveComm by Megvii Engine Team 2020-05-24 19:59:36 +0800
  • 26242d6b test(mgb/opr-mm): add collective_comm tests by Megvii Engine Team 2020-03-29 10:32:45 +0800
  • cd8ab9e3 test(mgb/opr-mm): add io_remote test by Megvii Engine Team 2020-03-29 11:56:09 +0800
  • 4e0054f7 fix(mgb/opr-mm): fix megray_helper thread safety by Megvii Engine Team 2020-05-14 14:48:57 +0800
  • 35d46dbb fix(mge/functional): simplify the api of add/remove_axis by Megvii Engine Team 2020-05-21 17:11:47 +0800
  • f7d8b516 feat(mge/functional): add smooth l1 loss by Megvii Engine Team 2020-05-18 18:01:34 +0800
  • 3c49d1d3 feat(mge/functional): add hinge loss by Megvii Engine Team 2020-05-18 17:58:16 +0800
  • dd8f3ffc test(distributed): replace python command in subprocess with `sys.executable` by Megvii Engine Team 2020-05-21 12:50:07 +0800
  • 7aa7a09b feat(mge/optimizer): add optimizer adadelta by Megvii Engine Team 2020-05-18 16:45:12 +0800
  • 205291a3 feat(mge/quantization): add histgram observer by Megvii Engine Team 2020-05-18 16:03:31 +0800
  • 7c4f1a38 feat(mge/quantization): add calibration support by Megvii Engine Team 2020-05-09 14:55:30 +0800
  • 5eca4da3 feat(mge/functional): add softplus function by Megvii Engine Team 2020-05-21 12:07:55 +0800
  • 855c49ca feat(mge/module): add linear quantization module by Megvii Engine Team 2020-05-13 09:54:13 +0800
  • 90107b6d feat(mge/cgtools): add network vistior interface with optional pruning by Megvii Engine Team 2020-05-11 10:02:14 +0000
  • 270b7488 feat(dnn/fallback): support mk4 fp32 im2col by Megvii Engine Team 2020-05-14 17:42:16 +0800
  • 45e2beea feat(mgb/gopt): add nchw4 optpass by Megvii Engine Team 2020-05-18 13:33:00 +0800
  • f2e1bb41 feat(mgb/opr): let more indexing ops support empty shape by Megvii Engine Team 2020-05-18 15:37:03 +0800
  • a4879fc6 feat(cmake/cross_build/host_build/windows): imp windows host build and cross build by Megvii Engine Team 2020-05-18 16:10:28 +0800
  • cdefe90e feat(dnn/fallback): support mk4 fp32 conv1x1 by Megvii Engine Team 2020-05-13 16:21:10 +0800
  • 980ebf2c feat(mge/module): add fused conv_bn qat approximate version by Megvii Engine Team 2020-05-16 16:44:38 +0800
  • 6972bfde fix(mge/distributed): fix input comp_graph of broadcast operator by Megvii Engine Team 2020-05-19 10:15:43 +0800
  • 66950a4f feat(dnn/arm): add nchw44 fp32 direct conv stride2 by Megvii Engine Team 2020-05-15 11:12:32 +0800
  • bfe945fb fix(dnn/arm): fix pooling printf datatype by Megvii Engine Team 2020-05-15 11:12:32 +0800
  • e28dc606 refactor(opr/dnn): support new MegDNN conv interface by Megvii Engine Team 2020-05-15 18:31:38 +0800
  • 0b320568 feat(dnn/naive): support weight preprocess interface in dnn by Megvii Engine Team 2020-05-15 13:59:57 +0800
  • b2f0ceb2 feat(dnn/naive): add convolution weight preprocess interface by Megvii Engine Team 2020-05-08 18:18:52 +0800
  • 9b908c02 refactor(opr/dnn): remove MegDNNOprInputsLayoutModifier by Megvii Engine Team 2020-05-15 17:46:40 +0800
  • 3adad485 feat(mge/optimizer): add optimizer adagrad by Megvii Engine Team 2020-05-18 14:37:36 +0800
  • 8be78b11 test(mge/optimizer): refactor the unittest of optimizer by Megvii Engine Team 2020-05-18 14:36:17 +0800