* Margaux dev.

* Support Pad, Activation, Mul operator for mnasnet

* Support document enhance model

* Apply add-init-cache-and-preemption patch

* Refine code

* Add APU int16 quantization and fix macro bugs

* Refine code
Co-authored-by: NYungChien Hsu <yungchien.hsu@mediatek.com>
Co-authored-by: NEric Chen <eric-yk.chen@mediatek.com>

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

Signed-off-by: Nlichao18 <lichao18@xiaomi.com>

Add init cache and preemption

Support mnasnet and new 5g library

Add GetModelSize API for model header

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

add input/output for dynamic lstm
use forward indexes for subsample

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

Revert "Merge branch 'refactor-python-scripts' into 'master'"

This reverts merge request !1200

fix type error

fix type warning

add TileOp to MaceTransposableDataFormatOps and reduce memcpy in TileOp

refactor structure of python scripts

N/A
Signed-off-by: NLuxuhui <luxuhui@xiaomi.com>

* Integrate APU

* fix code style issue

Summary: Reformat fp16 gemv, add FP16_MATMUL_WEIGHT

Differential Revision: https://phabricator.d.xiaomi.net/D144134

1. Change DataFormat from enum to enum class.

1. Auto transformation between data format, data type and memory type.
2. Add device placement optimization strategy.

add extract pooling and kaldi batchnorm

add time index info in extract pooling

add benchmar& test for kaldi components

add batch support for kaldi

change ifdefined to delay

support kaldi tdnn-lstm

support logsoftmax

add ifdefined op

add dynamic lstm docs

1. Support int32 input data type.
2. Support GatherV2 op
3. Add transpose to ExpandDim op.