v853NPU gaitset模型部署失败

carpediem LV 4

把训练好的gaitset.onnx转化为network_binary.nb模型后，在开发板上运行时，出现错误，如下图所示。

root@TinaLinux:~# vpm_run sample.txt
vpm_run sample.txt loop_run_count device_id
    sample.txt: to include one ore more network binary graph (NBG) data file resource.  See sample.txt for details.
loop_run_count: the number of loop run network.
device_id: specify this NBG runs device.
example: ./vpm_run sample.txt 1 1, specify the NBG runs on device 1.
         ./vpm_run sample.txt 1000, run this network 1000 times.

test started.

init vip lite, driver version=0x00010800...
[0xb6fbd560]vip_init[104],
The version of Viplite is: 1.8.0-0-AW-2022-04-21
vip lite init OK.

cid=0xee, device_count=1
  device[0] core_count=1
init test resources, batch_count: 1 ...
create/prepare networks ...
batch i=0, binary name: ./network_binary.nb
input 0 dim 44 64 100 1, data_format=2, quant_format=2, name=input[0], scale=0.003922, zero_point=0
ouput 0 dim 256 62 1 0, data_format=1, name=uid_1_out_0, none-quant
nbg name=./network_binary.nb
create network 1: 24090 us.
memory pool size=11265280byte
read golden file ./outputg.tensor
input 0 name: ./input.tensor
prepare network 0: 395732 us.
batch: 0, loop count: 1
start to run network=./network_binary.nb
[0xb6fbd560]gcvip_os_call_kernel[344], fail to ioctl vipcore, command[4]:CMD_WAIT, status=-1
[0xb6fbd560]gcvip_user_wait[484], failed to check status=-1
[0xb6fbd560]gcvip_capture_init[1240], catpure file name .//viplite_hang_capture_617de3_460_b6fbd560.log

后来观察转换过程中的模型结构，发现存在问题。

上面这个是简化onnx模型后的gaitset-sim.onnx的结构图

这个是导出nb模型之后保存的gaitset-simprj_fused.json的结构图。其中pooling之后的数据的结构为【44，32，0，32】，这个是不是有问题？
请问大家有解决这个问题的办法吗？
附上模型转换过程中的代码：

python -m onnxsim gaitset.onnx gaitset-sim.onnx --input-shape 1,100,64,44

pegasus import onnx --model gaitset-sim.onnx --output-data gaitset-sim.data --output-model gaitset-sim.json --outputs "feature"

pegasus generate inputmeta --model gaitset-sim.json --input-meta-output gaitset-sim_inputmeta.yml

pegasus generate postprocess-file --model gaitset-sim.json --postprocess-file-output gaitset-sim_postprocessmeta.yml

pegasus quantize --model gaitset-sim.json --model-data gaitset-sim.data --batch-size 1 --device CPU --with-input-meta gaitset-sim_inputmeta.yml --rebuild-all --model-quantize gaitset-sim.quantize --quantizer asymmetric_affine --qtype uint8

pegasus inference --model gaitset-sim.json --model-data gaitset-sim.data --batch-size 1 --dtype quantized --model-quantize gaitset-sim.quantize --device CPU --with-input-meta gaitset-sim_inputmeta.yml --postprocess-file gaitset-sim_postprocessmeta.yml

pegasus export ovxlib --model gaitset-sim.json --model-data gaitset-sim.data --dtype quantized --model-quantize gaitset-sim.quantize --batch-size 1 --save-fused-graph --target-ide-project 'linux64' --with-input-meta gaitset-sim_inputmeta.yml --output-path ovxilb/gaitset-sim/gaitset-simprj --pack-nbg-unify --postprocess-file gaitset-sim_postprocessmeta.yml --optimize "VIP9000PICO_PID0XEE" --viv-sdk ${VIV_SDK}

WhycanService LV 8

有问题，手动修改json模型结构试一下