使用pre_compile=True building model失败

[复制链接] · 发表于 2021-12-21 11:55:43

在rknn.build设置pre_compile=True后，报错。设置代码节选如下：

        # Build model

        print('--> Building model')

        ret = rknn.build(do_quantization=True, pre_compile=True, dataset='./dataset.txt')

        if ret != 0:

            print('build model failed.')

            exit(ret)

        print('done')
复制代码

报错结果如下：

    (venv) wenji-G3:~/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn$ python test.py

    --> Loading model

    done

    --> Building model

    W The RKNN Model generated can not run on simulator when pre_compile is True.

    make: gcc-linaro-6.3.1-2017.05-i686_aarch64-linux-gnugcc: Command not found

    make: *** [vnn_pre_process.o] Error 127

    /bin/sh: 1: /tmp/tmp0s82ra97/gen_nbg: not found

    mv: 无法获取'/home/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn/*.nb' 的文件状态(stat): 没有那个文件或目录

    mv: 无法获取'/home/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn/*.dat' 的文件状态(stat): 没有那个文件或目录

    E Generate nbg failed.

    W ----------------Warning(1)----------------

    E Catch exception when building RKNN model!

    E Traceback (most recent call last):

    E File "rknn/base/ovxconfiggenerator.py", line 193, in rknn.base.ovxconfiggenerator.generate_vx_config_from_files

    E File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 632, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator.generate

    E File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 583, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator._gen_special_case

    E File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 548, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator._gen_nb_file

    E AttributeError: 'CaseGenerator' object has no attribute 'nbg_graph_file_path'

    E During handling of the above exception, another exception occurred:

    E Traceback (most recent call last):

    E File "rknn/api/rknn_base.py", line 601, in rknn.api.rknn_base.RKNNBase.build

    E File "rknn/api/rknn_base.py", line 414, in rknn.api.rknn_base.RKNNBase._build

    E File "rknn/base/ovxconfiggenerator.py", line 195, in rknn.base.ovxconfiggenerator.generate_vx_config_from_files

    E File "rknn/api/rknn_log.py", line 312, in rknn.api.rknn_log.RKNNLog.e

    E ValueError: Generate nbg failed.

    build model failed.





复制代码

/tmp 目录下无 tmp0s82ra97/gen_nbg 请问这个目录文件如何生成的？需要自己添加吗？在sdk中有没有？
/home/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn/*.nb'及 /*.dat 两个文件也不存在，这两个文件需要获取吗？如何获取？

看了《Rockchip_Trouble_Shooting_RKNN_Toolkit_V1.4.0_CN.pdf》中关于rknn.build()设置pre_compile=True报错，其中写道需要确认

请确认:
1) 系统装有 gcc 编译工具链
2) 模型的名称只包含“字母”、“数字”、“_”,特别注意有些复制过来的模型名称
会自动加上”(1)”,这时会失败。也可以升级到 1.3.0 或之后的版本,由 RKNN Toolkit 自
动处理这些特殊字符。
如果不属于上述情况,可以尝试使用 export_rknn_precompile_model 接口从板子上导出
预编译模型。

系统为ubuntu18.04装有gcc，rknn-toolkit 1.6.0,

使用export_rknn_precompile_model的代码如下：

    print('--> init runtime')

    # ret = rknn.init_runtime()

    ret = rknn.init_runtime(target='rv1126', device_id='3166854b086a700f', rknn2precompile=True)

    if ret != 0:

        print('init runtime failed.')

        exit(ret)

    print('done')







    print('--> export pre-compile model')

    # Note:the rknn2precompile must be set True when call init_runtime

    ret = rknn.export_rknn_precompile_model('./test_pre_compile.rknn')

    if ret != 0:

        print('export pre-compile model failed.')

        exit(ret)





    rknn.release()

复制代码

仍然报上述错误，如下所示：

(venv) wenji-G3:~/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn$ python test.py

--> Loading model

done

--> Building model

W The RKNN Model generated can not run on simulator when pre_compile is True.

make: gcc-linaro-6.3.1-2017.05-i686_aarch64-linux-gnugcc: Command not found

make: *** [vnn_pre_process.o] Error 127

/bin/sh: 1: /tmp/tmp0s82ra97/gen_nbg: not found

mv: 无法获取'/home/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn/*.nb' 的文件状态(stat): 没有那个文件或目录

mv: 无法获取'/home/rv1109/external/rknn-toolkit/examples/darknet/yolov3_rknn/*.dat' 的文件状态(stat): 没有那个文件或目录

E Generate nbg failed.

W ----------------Warning(1)----------------

E Catch exception when building RKNN model!

E Traceback (most recent call last):

E File "rknn/base/ovxconfiggenerator.py", line 193, in rknn.base.ovxconfiggenerator.generate_vx_config_from_files

E File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 632, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator.generate

E File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 583, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator._gen_special_case

E File "rknn/base/RKNNlib/app/exporter/ovxlib_case/casegenerator.py", line 548, in rknn.base.RKNNlib.app.exporter.ovxlib_case.casegenerator.CaseGenerator._gen_nb_file

E AttributeError: 'CaseGenerator' object has no attribute 'nbg_graph_file_path'

E During handling of the above exception, another exception occurred:

E Traceback (most recent call last):

E File "rknn/api/rknn_base.py", line 601, in rknn.api.rknn_base.RKNNBase.build

E File "rknn/api/rknn_base.py", line 414, in rknn.api.rknn_base.RKNNBase._build

E File "rknn/base/ovxconfiggenerator.py", line 195, in rknn.base.ovxconfiggenerator.generate_vx_config_from_files

E File "rknn/api/rknn_log.py", line 312, in rknn.api.rknn_log.RKNNLog.e

E ValueError: Generate nbg failed.

build model failed.

复制代码

请问这种情况应该如何解决？

只看该作者 · 发表于 2021-12-21 12:31:37

make: gcc-linaro-6.3.1-2017.05-i686_aarch64-linux-gnugcc: Command not found

只看该作者 · 发表于 2021-12-21 17:47:03

已发现是编译器的问题，在环境里看到
export PATH=$PATH:~/rv1109/external/RKNPUTools/rknn-api/Linux/rknn_api_sdk/gcc-linaro-6.3.1-2017.05-i686_aarch64-linux-gnu/bin
export PATH=$PATH:~/rv1109/prebuilts/gcc/linux-x86/arm/gcc-arm-8.3-2019.03-x86_64-arm-linux-gnueabihf/bin
环境路径没有问题，在rv1109目录下搜索各种build.sh文件中，没有发现gcc-linaro-6.3.1-2017.05-i686_aarch64-linux-gnugcc这个路径，都是上边两条正确的路径，报错的编译器后缀带有gcc。请问我该如何解决这个问题？