 Adrian Lundell
|
2e5dcdc815
Use legacy Keras 2.0 for test data generation (#120)
|
1 rok temu |
|
Måns Nilsson
|
f0957f8ebc
Optimize convolution int8 generic MVE (#118)
|
2 lat temu |
|
Ryan OShea
|
6cc31fb36f
Int4 Depthwise performance improvement (#117)
|
2 lat temu |
|
Adrian Lundell
|
4b46c85b7a
Update mbed-os version (#113)
|
2 lat temu |
|
Måns Nilsson
|
20c92149b5
MVE Conv 1xN: Handle corner case (#111)
|
2 lat temu |
|
Måns Nilsson
|
72e1ebf623
Add non zero filter offset support for FC (#110)
|
2 lat temu |
|
Adrian Lundell
|
1e0f44c192
Align arm_vector_sum_s8 behaviour between default/MVE case (#107)
|
2 lat temu |
|
dependabot[bot]
|
5aeada78e1
Bump aws-actions/configure-aws-credentials from 4.0.1 to 4.0.2 (#105)
|
2 lat temu |
|
Ryan OShea
|
9eacdff489
Add dsp and mve support to transpose conv int8 (#103)
|
2 lat temu |
|
Adrian Lundell
|
2a999a2fd8
Add support for LSTM timing mode=False (#104)
|
2 lat temu |
|
Adrian Lundell
|
601d96c63a
Reimplement arm_lstm_unidirectional_s8 (#102)
|
2 lat temu |
|
Adrian Lundell
|
ffeca90436
Add grouped convolution to arm_convolve_s8 (#99)
|
2 lat temu |
|
Måns Nilsson
|
3b4e406b14
Correct internal compiler flagging (#98)
|
2 lat temu |
|
dependabot[bot]
|
90ffad8615
Bump aws-actions/configure-aws-credentials from 1.pre.node16 to 4.0.1 (#96)
|
2 lat temu |
|
dependabot[bot]
|
ba61675b9b
Bump actions/checkout from 3 to 4 (#97)
|
2 lat temu |
|
dependabot[bot]
|
ef0731e53e
Bump actions/setup-python from 4 to 5 (#95)
|
2 lat temu |
|
Vladimir Marchenko
|
e32a449aa0
Updated pack and doc build flows (#93)
|
2 lat temu |
|
Måns Nilsson
|
c9c8b3d49a
Add batched pooling support (#90)
|
2 lat temu |
|
Måns Nilsson
|
29c331e2f8
Rename include guards to align to naming convention (#91)
|
2 lat temu |
|
RyanOShea
|
040da18234
Add compiler optimization variable to CMake cache (#88)
|
2 lat temu |
|
ArmRyan
|
bfc54edb61
4-bit support for convolution (#85)
|
2 lat temu |
|
Måns Nilsson
|
edececa217
Add 4-bit weight support to depthwise conv (#83)
|
2 lat temu |
|
Måns Nilsson
|
ca476254fe
Add scalar version of transpose conv s8 (#82)
|
2 lat temu |
|
Adrian Lundell
|
d1c35d332c
Revert 2:3 read optimization for 1x1 convolution operator (#81)
|
2 lat temu |
|
Fotios Valasiadis
|
672a739a5a
Explicitly cast (void *) to (const int32_t *) (#76)
|
2 lat temu |
|
Adrian Lundell
|
6744ffd461
1x1_conv optimization improvements (#74)
|
2 lat temu |
|
Adrian Lundell
|
00cbf2ca16
Add optimized 4-bit fully connected operator + unit tests (#73)
|
2 lat temu |
|
Adrian Lundell
|
b4265c45f0
Add convolve_1x1_s8 MVE 2:3 read optimization. (#72)
|
2 lat temu |
|
Måns Nilsson
|
85164a8119
Fix pack generation (#71)
|
2 lat temu |
|
Måns Nilsson
|
58f1770576
MVE: Move kernel sums from core loop for FC and SVDF (#69)
|
2 lat temu |
