softmax算子支持

[ccw1996]

1. 添加softmax算子
2. 添加softmax测试用例

[ccw1996]

测试用例运行结果：

start to run register cpu allocator
AutoKernel plugin inited
function:autokernel_plugin_init executed
start init_tengine
init_tengine done
REPEAT COUNT= 1
Repeat [1] time 3.00 us per RUN. used 3 us
print output data
7.80134e-05 0.000212062 0.000576445 0.00156694 0.00425939 0.0115782 0.0314729 0.0855521 0.232555 0.632149

[lyuchuny3]

Hi, @ccw1996
你提交的softmax 执行 squeezenet网络后，结果出错。

cd autokernel_plugin
export DEBUG_INFO=1
./build/tests/tm_classification -n squeezenet

结果如下

runing AutoKernel im2col_conv ...
runing AutoKernel im2col_conv ...
[INFO]: runing Autokernel halide_avgpool...
[INFO]: runing Autokernel halide_softmax...
current 32.637 ms

Model name : squeezenet
tengine model file : models/squeezenet.tmfile
label file : models/synset_words.txt
image file : images/cat.jpg
img_h, imag_w, scale, mean[3] : 227 227 1 104.007 116.669 122.679

Repeat 1 times, avg time per run is 32.637 ms
max time is 32.637 ms, min time is 32.637 ms
--------------------------------------
116.6267 - "n04487081 trolleybus, trolley coach, trackless trolley"
103.5060 - "n02799071 baseball"
95.7461 - "n02112018 Pomeranian"
91.3066 - "n03662601 lifeboat"
86.8204 - "n04487394 trombone"
--------------------------------------

正确结果是

--------------------------------------
0.2732 - "n02123045 tabby, tabby cat"
0.2676 - "n02123159 tiger cat"
0.1810 - "n02119789 kit fox, Vulpes macrotis"
0.0818 - "n02124075 Egyptian cat"
0.0724 - "n02085620 Chihuahua"
--------------------------------------

[ccw1996]

@lyuchuny3 十分抱歉，之前只验证了自己的测试样例，疏忽了。
现在已经修改完成，使用了tm_classification和test_softmax进行测试，均能够得到正确结果。