Megvii Engine Team
328fb36f17
feat(mgb/opr-mm): add Scatter, Gather, AllToAll oprs
GitOrigin-RevId: f75169ecd6
5 years ago
Megvii Engine Team
3f51a6a05b
build(third_party): Update megray version
5 years ago
Megvii Engine Team
285d70cb66
fix(build): find_package(Threads) if not already done
GitOrigin-RevId: 10ecb51947
5 years ago
Megvii Engine Team
786afef461
feat(build): install CMake config module and pkg-config descriptor
Also, upgrade to CMake 3.13.
The commit also contains significant refactors, as otherwise it is not
possible to properly export target `megengine` to
MegEngine-targets.cmake:
1. Optionally use system provided Flatbuffers.
2. Optionally use system provided MKL-DNN (Tested with Debian).
3. Refactor megbrain and megdnn targets into object libraries.
4. Set different path in BUILD_INTERFACE and INSTALL_INTERFACE of
various target_include_directories.
5. Specify PUBLIC/PRIVATE on various target_link_libraries.
GitOrigin-RevId: df118a879e
5 years ago
Megvii Engine Team
4d35397bdf
fix(dnn/fallback): fix conv1x1/im2col usable and fuse-conv-bias get fp32xfp32-->qint8 error
GitOrigin-RevId: 5a3bfedd8a
5 years ago
Megvii Engine Team
12dc36a6ab
feat(mgb/gopt): add interface to reproducible
GitOrigin-RevId: f341bea40b
5 years ago
Megvii Engine Team
cc4e1dfd7c
feat(mgb/compnode): expose mem_status and try_coalesce_all_free_memory to python
GitOrigin-RevId: 0ec2d556b1
5 years ago
Megvii Engine Team
7bcead7561
ci(docker): add build-arg option to setup yum and pip mirrors
GitOrigin-RevId: bb7aa875da
5 years ago
Megvii Engine Team
6b2760dd72
feat(dnn/fallback): add float32 nchw44 fuse packb 3x3 s2
GitOrigin-RevId: 3b664bb4f5
5 years ago
Megvii Engine Team
7aeb4f6ca7
fix(mge/optimizer): use static key to avoid mem leak
GitOrigin-RevId: 85298084a3
5 years ago
Megvii Engine Team
7a0c7ef45c
feat(mge/module): add module for extern-c-opr
GitOrigin-RevId: a2d9fa067a
5 years ago
Megvii Engine Team
09d2b7c3fe
fix(core): make the semantics of instance id clear and correct
GitOrigin-RevId: 2232195c50
5 years ago
Megvii Engine Team
2b4b4d66d9
feat(dnn/fallback): add aarch64 mk4 dot 3x3 s1 fuse packb
GitOrigin-RevId: 3e69878d8d
5 years ago
Megvii Engine Team
a1677d7aa9
feat(dnn/arm_common): add fp32 gevm
GitOrigin-RevId: 4d348bbb34
5 years ago
Megvii Engine Team
5d950063cf
feat(dnn): refactor dot gemv for both aarch64 and aarch32
GitOrigin-RevId: 2b98867e45
5 years ago
Megvii Engine Team
53c288a304
fix(dnn/cuda): fix topk grid oversize
GitOrigin-RevId: d3c811a034
5 years ago
Megvii Engine Team
124767b4f8
fix(dnn/fallback): fix mk4_dot test after remove mk4_dot_8x6x4 matmul
GitOrigin-RevId: e3a12cf9b3
5 years ago
Megvii Engine Team
34659c2ea4
fix(mgb/dnn): remove armv7 matmul mk4dot block 8x6
GitOrigin-RevId: 4c746ef228
5 years ago
Megvii Engine Team
48ac1e1abd
feat(dnn/fallback): delete nopack onlypacka noneed datatype,and add
im2co and conv1x1 mk4_dot support
GitOrigin-RevId: 096b16a3ab
5 years ago
Megvii Engine Team
3117bfb738
fix(dnn/arm): nchw44 direct int8 support 8832
GitOrigin-RevId: 696fa05d94
5 years ago
Megvii Engine Team
4e0c9ad3c6
feat(mgb/external): extern-c-opr dumper and loader for MACE
GitOrigin-RevId: bfe5420b1d
5 years ago
Megvii Engine Team
ca52a93e9f
fix(mge/quant): fix init value of histogram observer
GitOrigin-RevId: 9c8caa5f8f
5 years ago
Megvii Engine Team
9f352b1c45
feat(megbrain/dnn): add indexing remap int32 for naive and cuda
GitOrigin-RevId: 5f66d51de4
5 years ago
Megvii Engine Team
5dbf218d19
feat(dnn/x86): add sse 8816 matmul
GitOrigin-RevId: ed8d9ee5db
5 years ago
Megvii Engine Team
25b6a13148
feat(dnn/x86): add x86 avx2 8x8x16 matmul
GitOrigin-RevId: d2172c50b2
5 years ago
Megvii Engine Team
273f891b55
fix(mgb/gopt): fix run-time winograd-transform and nchwxx error
GitOrigin-RevId: aca796f17d
5 years ago
Megvii Engine Team
02abc36ea6
fix(mbg/arm_common): fix nchw44-dot misc issue
GitOrigin-RevId: f870ad964c
5 years ago
Megvii Engine Team
9ed3882a94
fix(opr/dnn): fix winograd fast run mismatch
GitOrigin-RevId: d308085b9f
5 years ago
Megvii Engine Team
18be23f328
fix(mbg/gopt): fix nchwxx gopt with no fuse conv_bias and winograd
fast-run
GitOrigin-RevId: 49ccbdf2d4
5 years ago
Megvii Engine Team
38f7cbd9aa
fix(mge/module): fix redundant recursion in `train()`
GitOrigin-RevId: 6b3566930b
5 years ago
Megvii Engine Team
5c2323529d
test(mge/quantization): add `quantize_disabled` related test
GitOrigin-RevId: f62ba600c5
5 years ago
Megvii Engine Team
ab91302515
feat(mge/quantization): add `quantize_disabled` attribute in Module
GitOrigin-RevId: f108f03c5a
5 years ago
Megvii Engine Team
f4ead78852
feat(mgb): static allocation with given padding
GitOrigin-RevId: fdf2de8ad6
5 years ago
Megvii Engine Team
575a6dca9f
ci(docker): change object files permission
GitOrigin-RevId: e388f4b57e
5 years ago
Megvii Engine Team
08ac685e3a
feat(mge/functional): add logsumexp
GitOrigin-RevId: e7ef50e9ec
5 years ago
Megvii Engine Team
65ec4f7c26
fix(ci): fix test timeout
GitOrigin-RevId: 875fc613cf
5 years ago
Megvii Engine Team
ea6bfe6cd9
fix(dnn/cuda-stub): simplify and use proper search paths
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: ed43cab8c8
5 years ago
Megvii Engine Team
01092feb9b
feat(mgb): add PackAllReducePass
GitOrigin-RevId: 59c1b45393
5 years ago
Megvii Engine Team
c7e6c658fd
refactor(mge/distribute): use is_root (and rank) in stead of rank and root at collective comm
GitOrigin-RevId: dccdb71553
5 years ago
Megvii Engine Team
ff308e3b62
feat(mgb/comp_node): generate uid for cuda comp node
GitOrigin-RevId: 34fa5a2fb6
5 years ago
Megvii Engine Team
32c86211ee
fix(dnn/cuda): enable cuda algos for nchw quantized
GitOrigin-RevId: 4d1e167b86
5 years ago
Megvii Engine Team
b8d8886e35
refactor(mge/tensor): combine Dict and TensorDict
GitOrigin-RevId: 6b6c03c04b
5 years ago
Megvii Engine Team
7751a0676e
docs(mge/tensor): add advanced index related docs
GitOrigin-RevId: 31735ddac4
5 years ago
Megvii Engine Team
7b0dbe6af8
fix(dnn/arm): fix stride 1 support for int8 nchw_nchw44
GitOrigin-RevId: 9d718eb7a4
5 years ago
Megvii Engine Team
198f3eb5f6
fix(dnn/arm): fix fp32 nchw44 direct workspace bug
GitOrigin-RevId: 6ee433b02c
5 years ago
Megvii Engine Team
49fdddef8d
fix(gopt): fix reorder arith chain pass
GitOrigin-RevId: d3257ac43a
5 years ago
Megvii Engine Team
6742a58b7e
fix(quant): observer do not use cond_take
GitOrigin-RevId: a954814bcb
5 years ago
Megvii Engine Team
9e876203b5
feat(dnn): add int8 direct conv dot nchw44
GitOrigin-RevId: 31830ba7a4
5 years ago
Megvii Engine Team
09ceaaaecf
fix(dnn/arm): stride1 support for nchw_nchw44 fp32 conv
GitOrigin-RevId: 744c5db3dc
5 years ago
Megvii Engine Team
50db9b84c2
fix(gopt): fix paramfuse if the endpoint is const
GitOrigin-RevId: f666f6d700
5 years ago