提交 · 34659c2ea4a2f1134268b9e91bff01a990f29e63 · MegEngine 天元 / MegEngine

19 6月, 2020 40 次提交
- M
  fix(mgb/dnn): remove armv7 matmul mk4dot block 8x6 · 34659c2e
  由 Megvii Engine Team 提交于 6月 12, 2020
```
GitOrigin-RevId: 4c746ef22895ebc6ab4298c66f71e239b437cc69
```
  34659c2e
- M
  feat(dnn/fallback): delete nopack onlypacka noneed datatype,and add · 48ac1e1a
  由 Megvii Engine Team 提交于 5月 29, 2020
```
im2co and conv1x1 mk4_dot support

GitOrigin-RevId: 096b16a3abb8fa259db77e792bd286dacf3fd8c3
```
  48ac1e1a
- M
  fix(dnn/arm): nchw44 direct int8 support 8832 · 3117bfb7
  由 Megvii Engine Team 提交于 6月 08, 2020
```
GitOrigin-RevId: 696fa05d943b28fcec3a236bb8518fb255eae9db
```
  3117bfb7
- M
  feat(mgb/external): extern-c-opr dumper and loader for MACE · 4e0c9ad3
  由 Megvii Engine Team 提交于 6月 08, 2020
```
GitOrigin-RevId: bfe5420b1de97eae9ee943c4d3adaf652d795e03
```
  4e0c9ad3
- M
  fix(mge/quant): fix init value of histogram observer · ca52a93e
  由 Megvii Engine Team 提交于 6月 12, 2020
```
GitOrigin-RevId: 9c8caa5f8f2b02e881601a9aa6ff30f654e30a07
```
  ca52a93e
- M
  feat(megbrain/dnn): add indexing remap int32 for naive and cuda · 9f352b1c
  由 Megvii Engine Team 提交于 6月 05, 2020
```
GitOrigin-RevId: 5f66d51de4751d77fc05e2849388fc6dfbae4a53
```
  9f352b1c
- M
  feat(dnn/x86): add sse 8816 matmul · 5dbf218d
  由 Megvii Engine Team 提交于 6月 12, 2020
```
GitOrigin-RevId: ed8d9ee5db3342ec534dc2e8a38b8613b8733b2b
```
  5dbf218d
- M
  feat(dnn/x86): add x86 avx2 8x8x16 matmul · 25b6a131
  由 Megvii Engine Team 提交于 6月 12, 2020
```
GitOrigin-RevId: d2172c50b244a0683b00710a88bc507d02a9734f
```
  25b6a131
- M
  fix(mgb/gopt): fix run-time winograd-transform and nchwxx error · 273f891b
  由 Megvii Engine Team 提交于 6月 10, 2020
```
GitOrigin-RevId: aca796f17defd041802e926e9f0742b02dd48de4
```
  273f891b
- M
  fix(mbg/arm_common): fix nchw44-dot misc issue · 02abc36e
  由 Megvii Engine Team 提交于 6月 11, 2020
```
GitOrigin-RevId: f870ad964c075fe55fb1c7ca131680873bec61eb
```
  02abc36e
- M
  fix(opr/dnn): fix winograd fast run mismatch · 9ed3882a
  由 Megvii Engine Team 提交于 6月 05, 2020
```
GitOrigin-RevId: d308085b9fe16f8aae874346a08f55428a85bb76
```
  9ed3882a
- M
  fix(mbg/gopt): fix nchwxx gopt with no fuse conv_bias and winograd · 18be23f3
  由 Megvii Engine Team 提交于 6月 01, 2020
```
fast-run

GitOrigin-RevId: 49ccbdf2d43229f3883af91f9f9641f695e2a799
```
  18be23f3
- M
  fix(mge/module): fix redundant recursion in `train()` · 38f7cbd9
  由 Megvii Engine Team 提交于 6月 12, 2020
```
GitOrigin-RevId: 6b3566930b72c56b571debea1e1901745b0e7cdc
```
  38f7cbd9
- M
  test(mge/quantization): add `quantize_disabled` related test · 5c232352
  由 Megvii Engine Team 提交于 6月 12, 2020
```
GitOrigin-RevId: f62ba600c537f60e1229b852b97cd67897bf236d
```
  5c232352
- M
  feat(mge/quantization): add `quantize_disabled` attribute in Module · ab913025
  由 Megvii Engine Team 提交于 6月 11, 2020
```
GitOrigin-RevId: f108f03c5a5ac2142d5cb7bb9c24cbe43dce6fe6
```
  ab913025
- M
  feat(mgb): static allocation with given padding · f4ead788
  由 Megvii Engine Team 提交于 5月 26, 2020
```
GitOrigin-RevId: fdf2de8ad6f767bf3d0c4f3ae9287bbd77b70c16
```
  f4ead788
- M
  ci(docker): change object files permission · 575a6dca
  由 Megvii Engine Team 提交于 6月 11, 2020
```
GitOrigin-RevId: e388f4b57e46299fd7929ef4afba8b04ba6466ab
```
  575a6dca
- M
  feat(mge/functional): add logsumexp · 08ac685e
  由 Megvii Engine Team 提交于 6月 10, 2020
```
GitOrigin-RevId: e7ef50e9ece3ba6f71f5fe23c9a81ce9ccd0dc60
```
  08ac685e
- M
  fix(ci): fix test timeout · 65ec4f7c
  由 Megvii Engine Team 提交于 6月 10, 2020
```
GitOrigin-RevId: 875fc613cfcdd64ee4da0727725d9226f1363faa
```
  65ec4f7c
- M
  fix(dnn/cuda-stub): simplify and use proper search paths · ea6bfe6c
  由 Megvii Engine Team 提交于 5月 28, 2020
```
Removed the `access()` call before `dlopen()`.
It was copy-pasted from the opencl-stub, does not make sense here, and
prevents `dlopen()` from loading `libcuda.so` from non-default path.

Updated the name of the library providing CUDA Driver API on different
platforms, these are harvested from the following file in a CUDA
install:
samples/6_Advanced/matrixMulDynlinkJIT/cuda_drvapi_dynlink.c
GitOrigin-RevId: ed43cab8c8b4a9b9f8baaa958e8b5123b72dd179
```
  ea6bfe6c
- M
  feat(mgb): add PackAllReducePass · 01092feb
  由 Megvii Engine Team 提交于 6月 01, 2020
```
GitOrigin-RevId: 59c1b4539340f7bf1b9475255de665b1842accf1
```
  01092feb
- M
  refactor(mge/distribute): use is_root (and rank) in stead of rank and root at collective comm · c7e6c658
  由 Megvii Engine Team 提交于 5月 29, 2020
```
GitOrigin-RevId: dccdb715533576db97bbc21fe61e640141c1e8b6
```
  c7e6c658
- M
  feat(mgb/comp_node): generate uid for cuda comp node · ff308e3b
  由 Megvii Engine Team 提交于 6月 08, 2020
```
GitOrigin-RevId: 34fa5a2fb613de80ae14034ff4337ffd2ec3f3e3
```
  ff308e3b
- M
  fix(dnn/cuda): enable cuda algos for nchw quantized · 32c86211
  由 Megvii Engine Team 提交于 6月 04, 2020
```
GitOrigin-RevId: 4d1e167b86764ea18a0ea45e58491428b778aa74
```
  32c86211
- M
  refactor(mge/tensor): combine Dict and TensorDict · b8d8886e
  由 Megvii Engine Team 提交于 6月 08, 2020
```
GitOrigin-RevId: 6b6c03c04b7c97c29d30831e07c21e8afd3c3f40
```
  b8d8886e
- M
  docs(mge/tensor): add advanced index related docs · 7751a067
  由 Megvii Engine Team 提交于 6月 08, 2020
```
GitOrigin-RevId: 31735ddac487826aa604c2a2aec2d72b572fa609
```
  7751a067
- M
  fix(dnn/arm): fix stride 1 support for int8 nchw_nchw44 · 7b0dbe6a
  由 Megvii Engine Team 提交于 6月 03, 2020
```
GitOrigin-RevId: 9d718eb7a4dae3c2724ea07ba2b639fbfb319f78
```
  7b0dbe6a
- M
  fix(dnn/arm): fix fp32 nchw44 direct workspace bug · 198f3eb5
  由 Megvii Engine Team 提交于 6月 04, 2020
```
GitOrigin-RevId: 6ee433b02c31c40c154b9d22ad9116fd9afc38a7
```
  198f3eb5
- M
  fix(gopt): fix reorder arith chain pass · 49fdddef
  由 Megvii Engine Team 提交于 6月 04, 2020
```
GitOrigin-RevId: d3257ac43a69d37176896acfebaf1897ca213f05
```
  49fdddef
- M
  fix(quant): observer do not use cond_take · 6742a58b
  由 Megvii Engine Team 提交于 6月 09, 2020
```
GitOrigin-RevId: a954814bcbbd81736da5dd383d9492efc1ef8ed1
```
  6742a58b
- M
  feat(dnn): add int8 direct conv dot nchw44 · 9e876203
  由 Megvii Engine Team 提交于 5月 21, 2020
```
GitOrigin-RevId: 31830ba7a49f7c0b9fb3f011e09f934601a825a0
```
  9e876203
- M
  fix(dnn/arm): stride1 support for nchw_nchw44 fp32 conv · 09ceaaae
  由 Megvii Engine Team 提交于 6月 03, 2020
```
GitOrigin-RevId: 744c5db3dc3a867d1577f1c870e47472945234f5
```
  09ceaaae
- M
  fix(gopt): fix paramfuse if the endpoint is const · 50db9b84
  由 Megvii Engine Team 提交于 5月 27, 2020
```
GitOrigin-RevId: f666f6d70037debbff34551149d04b0bd8c256f4
```
  50db9b84
- M
  fix(mge/function): do not deeply copy saved tensor in Function · 35bc0e1f
  由 Megvii Engine Team 提交于 6月 01, 2020
```
GitOrigin-RevId: 3c89d1ceaa9f7338b167f80936549a4c364061de
```
  35bc0e1f
- M
  fix(core): fix memory defragmenter · 47377c7b
  由 Megvii Engine Team 提交于 6月 05, 2020
```
GitOrigin-RevId: e883be8b5ce8e6543838a5f84674bbe5791017ff
```
  47377c7b
- M
  fix(mbg/gopt): fix nchw44-dot channel wise trans to nchw44 · f56f187f
  由 Megvii Engine Team 提交于 6月 04, 2020
```
GitOrigin-RevId: aa2059a79601822131a29f8e23f6808847ca62c3
```
  f56f187f
- M
  feat(mgb/plugin): add param json func for indexing oprs · af29fcb2
  由 Megvii Engine Team 提交于 6月 01, 2020
```
GitOrigin-RevId: b5becbbc028b16e4e1dcf313a1acf1c8b5bb11d8
```
  af29fcb2
- M
  fix(mge/sdk): fix comp_node bug in dump_with_testcast_mge · 62753c4d
  由 Megvii Engine Team 提交于 6月 03, 2020
```
GitOrigin-RevId: 26a8dc50b8fb63e4cbec45785cb49aa74290f8d3
```
  62753c4d
- M
  fix(dnn/cuda): fix FuseConvBiasWithZ pass for HSwish activation · f1c86606
  由 Megvii Engine Team 提交于 6月 04, 2020
```
GitOrigin-RevId: b290469cb1993214c01cab823864476778963f70
```
  f1c86606
- M
  fix(mge/functional): fix scatter doctest failed for GPU platform issue · adfa4688
  由 Megvii Engine Team 提交于 6月 04, 2020
```
GitOrigin-RevId: b5f92c39dd46f13aee07e9cad441e017069aaa3a
```
  adfa4688

MegEngine 天元 / MegEngine 1 年多 前同步成功

MegEngine 天元 / MegEngine
1 年多前同步成功