Megvii Engine Team
dbd9483993
feat(dnn,src,imperative): add groupnorm op
GitOrigin-RevId: de3c3d10e5
2 years ago
Megvii Engine Team
f444d4fe4d
feat(dnn,imperative): region restricted conv support groups=1 even if
depthwise
GitOrigin-RevId: 950d2f4889
2 years ago
Megvii Engine Team
6db4620e6d
feat(dnn): fix wgrad rrconv for compute capability
GitOrigin-RevId: ba8792d7a9
2 years ago
Megvii Engine Team
4e9b1c4eee
feat(dnn): add rrconv wgrad, support int32 and uint8 region mask
GitOrigin-RevId: 0da9b3bca8
2 years ago
Megvii Engine Team
977c207171
feat(dnn): add RegionRestrictedConv DGRAD support int32 and uint8
GitOrigin-RevId: 814b8a83f8
2 years ago
Megvii Engine Team
543c9b77a8
feat(dnn): add RegionRestrictedConv cuda
GitOrigin-RevId: b9f2d34a13
2 years ago
Megvii Engine Team
58b682ca00
feat(dnn/cuda): add naive bmm
GitOrigin-RevId: 4ba4b22e40
3 years ago
Megvii Engine Team
669816e291
feat(dnn): warpperspective support multi src input
GitOrigin-RevId: 8a4789852e
2 years ago
Megvii Engine Team
d2a1905ad5
Revert "feat(mgb): add cumprod opr"
This reverts commit 3436c3bdaa
.
GitOrigin-RevId: 95ab3d1aa7
2 years ago
Megvii Engine Team
49e14f87b5
feat(mgb): add cumprod opr
GitOrigin-RevId: 3436c3bdaa
3 years ago
Megvii Engine Team
c49d3070ba
refactor(imperative/ops): extends DnnOprCaller with template
GitOrigin-RevId: 402cba209a
3 years ago
Megvii Engine Team
247e2f59a4
feat(mgb/dnn): add modes that the output type is bool in elemwise
GitOrigin-RevId: fd0134fca2
3 years ago
Megvii Engine Team
7b17c1180e
refactor(dnn): make cudnn_frontend work
GitOrigin-RevId: f089f93494
3 years ago
Megvii Engine Team
35e9cc9845
feat(dnn/cuda): add cudnn frontend api
GitOrigin-RevId: 9b18a57893
3 years ago
Megvii Engine Team
0d7ace15c8
fix(mgb/dnn): suport fp16 for resize nhwc
GitOrigin-RevId: bb04d2a801
3 years ago
Megvii Engine Team
b55942a94d
feat(dnn/naive/norm,-dnn/cuda/norm,-dnn/test/norm): add norm dnn opr,
fwd only
GitOrigin-RevId: 989474168d
3 years ago
Megvii Engine Team
bbafe69974
feat(dnn): add elemwise COND_LT_MOV
GitOrigin-RevId: 444cd6825a
3 years ago
Megvii Engine Team
98b5ee78c1
feat(mge/dnn): add lamb optimizer
GitOrigin-RevId: 5a27157456
3 years ago
Megvii Engine Team
c2e9860feb
chore(license): remove all license in file header
GitOrigin-RevId: a0e31247a6
3 years ago
Megvii Engine Team
70209667e8
fix(dnn/test): fix some bug when force_deduce_layout is off
GitOrigin-RevId: d7ccc397df
3 years ago
Megvii Engine Team
7dc347697a
feat(dnn/cuda): add typecvt uint16
GitOrigin-RevId: d1368c414e
3 years ago
Megvii Engine Team
4c0bff1dba
refactor(megdnn): refactor TEGRA_X1/X2 macro
GitOrigin-RevId: 1aa78712c6
3 years ago
Megvii Engine Team
758549b936
feat(megengine): support tx2
GitOrigin-RevId: d1175a1f4a
3 years ago
Megvii Engine Team
b6ad457269
feat(cuda): support int1 simplewq conv
GitOrigin-RevId: 9c37c41bc7
3 years ago
Megvii Engine Team
fd6f8e58b0
feat(mgb/dtype): add dtype qint1
GitOrigin-RevId: abe9fb68b1
3 years ago
Megvii Engine Team
87de704a46
feat(gopt): fuse conv h_swish
GitOrigin-RevId: a3d12991fb
3 years ago
Megvii Engine Team
d7b0994a3e
feat(cuda): add fp16 compute 16 kernel
GitOrigin-RevId: e03435be02
3 years ago
Megvii Engine Team
8a2e92bd6c
refactor(cuda): depthwish large kernel
GitOrigin-RevId: dade8710b4
3 years ago
Megvii Engine Team
6b8a69d5b6
feat(cuda): float16 depthwise large kernel conv compute fp32
GitOrigin-RevId: 3050d48f26
3 years ago
Megvii Engine Team
bc385b5374
feat(cuda): support float16 depthwise large kernel conv
GitOrigin-RevId: fdc1b15fbc
3 years ago
Megvii Engine Team
7d2063e35a
perf(cuda): speedup conv backward data with small feature map and large filter size
GitOrigin-RevId: 85592bca6b
3 years ago
Megvii Engine Team
72403e8929
perf(cuda): speedup chanwise conv with small feature map and large filter size
GitOrigin-RevId: e65b2ce856
3 years ago
Megvii Engine Team
ab6d12caff
feat(mge): add conv padding mode
GitOrigin-RevId: 147ced856e
3 years ago
Megvii Engine Team
47fe766310
feat(dnn/cuda): add implicit bmm kernels for large kernel depthwise convolution backward filter opr
GitOrigin-RevId: 932e7689e8
3 years ago
Megvii Engine Team
6cefabe734
fix(dnn/cuda): fix ci
GitOrigin-RevId: 8267e5f9dd
3 years ago
Megvii Engine Team
888f4e46ae
feat(dnn/cuda): add implicit bmm large kernel dwconv2d dgrad kernels
GitOrigin-RevId: fcb7974d62
3 years ago
Megvii Engine Team
08d8635ff5
feat(dnn/cuda): add implicit bmm large kernel dwconv2d fprop impl
GitOrigin-RevId: feb09ebb58
3 years ago
Megvii Engine Team
95ac055538
feat(dnn,mgb,imperative): add diag opr implement
GitOrigin-RevId: 43016ffa2b
3 years ago
Megvii Engine Team
cbbca5fb10
feat(mge): add softmax op use cudnn api
GitOrigin-RevId: 7734ebf8c4
3 years ago
Megvii Engine Team
82be0aaced
test(dnn): fix compute capability requirement for NCHWX test
GitOrigin-RevId: d2f8022be1
3 years ago
Megvii Engine Team
1999307015
feat(mgb/opr): add dropout kernel
GitOrigin-RevId: d248bd2005
3 years ago
Megvii Engine Team
a93741815b
feat(mgb/opr): add layernorm forward and backward kernel
GitOrigin-RevId: 0cd484e753
3 years ago
Megvii Engine Team
2696e4efaa
feat(dnn): add float16 for remap backward
GitOrigin-RevId: 0263030051
3 years ago
Megvii Engine Team
11d75fecb5
feat(dnn/check_non_finite): add batch check_non_finite
GitOrigin-RevId: e108133282
3 years ago
Megvii Engine Team
ba2f0c2e48
fix(dnn/cuda): fix cudnn_conv algo of conv_bias opr for fp16 add z cases
GitOrigin-RevId: b29b009de0
3 years ago
Megvii Engine Team
c85631aa77
feat(dnn): use ref ptr interface for all backends
GitOrigin-RevId: f65feae5cc
3 years ago
Megvii Engine Team
89186edc5d
fix(dnn): correct reduce/argmxx/fakequant calculation with nan
GitOrigin-RevId: 7e78bdae91
3 years ago
Megvii Engine Team
68cdabd288
feat(opr): indexing_multi_axis_vec support nd index
GitOrigin-RevId: 07b1248bdc
3 years ago
Megvii Engine Team
9b4cd92ba3
fix(mgb/dnn): fix cudnnConvBiasActivation crash on nchw32 int8 with oc > 256
GitOrigin-RevId: 20c0b90575
3 years ago
Megvii Engine Team
10af44abba
fix(dnn/cuda): fix cudnn conv impl for nchw4_nchw hybrid layout
the conv_bias algo *_IMPLICIT_GEMM in cudnn less than 8.0.0 is disabled due to the incorrect result for int8x4->f32 configs
GitOrigin-RevId: 7cc52d0a85
3 years ago