当运行六角nnlib中的独立graph_app时，Shell会被卡住

Question

当我发出此命令时，Shell没有响应。

/data/local/graph_app --flag 299 299 3 1 0 0 1 NULL 0 1 0 inputfile /data/local/tmp/img_299x299.bmp

从助手optargs获得描述信息

Usage: testapp [--flag flagopt] [inputfile [inputfile...]]
           flag name      type   default  function
              height       int         0  Height of the input data. 0 == autodetect-square
               width       int         0  Width of the input data. 0 == autodetect-square
               depth       int         3  Depth of the input data
               iters       int         1  Number of times to run each input
            perfdump       int         0  Generate performance dump
                 pmu       int         0  Get Performance Monitor Unit information
         elementsize       int         1  Element Size (uint8==1,float==4)
       layer_reorder    string      NULL  Reorder depth layers. ("210" changes RGB to BGR)
       pprint_floats       int         0  Pretty-Print output as floats
     pprint_imagenet       int         1  Pretty-print output, getting top 5 values and use imagenet categories
               debug       int         0  Debug verbosity level. Higher numbers get more verbosity

我错过了什么吗，让我知道，我使用graphinit_med.c只是为了检查它的工作，没有关于这个模型的描述。

谢谢,

Answer 1

没有关于使用独立的graph_app的文档，在遍历代码使其正常工作之后：

data/hvx_tf/graph_app --height 299 --width 299 --depth 3 --iters 1 --perfdump 0 --pmu 0 --elementsize 1 --pprint_floats 0 --pprint_imagenet 1 --debug 0 /data/local/tmp/keyboard_299x299.dat

>> Generate *.dat from *.jpg using `./scripts/imagedump.py`

您仍然可以看到以下日志中的警告：

return value from dspCV_initQ6() : 0 
const node 1000b success
const node 1000c success
const node 1000d success
const node 1000e success
const node 1000f success
const node 10010 success
const node 10011 success
const node 10012 success
const node 10250 success
nn @ fc72cf80: id=0x0 debug_level=0
node @ fc733970: id=0x1000b type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733a20: id=0x1000c type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733a70: id=0x1000d type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733b20: id=0x1000e type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733c20: id=0x1000f type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733c70: id=0x10010 type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733d30: id=0x10011 type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733e20: id=0x10012 type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733e70: id=0x10250 type=0x3(Const) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733ec0: id=0x1024a type=0x0(INPUT) n_inputs=0 n_outputs=1 padding=0(WHATEVER)
node @ fc733f60: id=0x1024b type=0xe(Flatten) n_inputs=2 n_outputs=1 padding=0(WHATEVER)
node @ fc734040: id=0x1024c type=0x29(Min_f) n_inputs=2 n_outputs=1 padding=0(WHATEVER)
node @ fc734120: id=0x1024d type=0x2b(Max_f) n_inputs=2 n_outputs=1 padding=0(WHATEVER)
node @ fc734200: id=0x1024e type=0x2d(Quantize) n_inputs=3 n_outputs=3 padding=0(WHATEVER)
node @ fc734350: id=0x1024f type=0xf(QuantizedConv2d_8x8to32) n_inputs=7 n_outputs=3 padding=2(VALID)
node @ fc7344d0: id=0x10251 type=0x13(QuantizeDownAndShrinkRange_32to8) n_inputs=3 n_outputs=3 padding=0(WHATEVER)
node @ fc734620: id=0x10252 type=0x23(QuantizedBiasAdd_8p8to32) n_inputs=6 n_outputs=3 padding=0(WHATEVER)
node @ fc734790: id=0x10253 type=0x13(QuantizeDownAndShrinkRange_32to8) n_inputs=3 n_outputs=3 padding=0(WHATEVER)
node @ fc7348e0: id=0x10254 type=0x15(QuantizedRelu_8) n_inputs=3 n_outputs=3 padding=0(WHATEVER)
node @ fc734a30: id=0x10442 type=0x2f(Dequantize) n_inputs=3 n_outputs=1 padding=0(WHATEVER)
node @ fc734b20: id=0x1044d type=0x1(OUTPUT) n_inputs=1 n_outputs=0 padding=0(WHATEVER)
21 nodes total.
Init graph done.Prepare fc72cf80 success!
Using </data/local/tmp/keyboard_299x299.dat>
filesize=268203 elementsize=1 height=299 width=299 depth=3
Run!
sum=37845659
Executing!
**execute got err: -1**
hexagon/ops/src/op_output.c:58:output 0 too small
output size=4096
Rank,Softmax,index,string
0,303036629674309094288042513882152960.000000,575,pick
1,303036292954618408664607741320364032.000000,461,terrapin
2,303036292954618408664607741320364032.000000,445,electric ray
3,79327539388858010780491752432205824.000000,833,bulletproof vest
4,78902425827607254052570293577187328.000000,936,volleyball
AppReported: 4294967296

我将更新答案，一旦我得到独立的应用程序预测样本图像的最高精度。