瑞芯微M1808开发板上运行目标检测提示Check failed: (device_names.size() > 0): 0!>0 No device specified.

[ysjhlnu]

• 标题：瑞芯微M1808开发板上运行目标检测提示Check failed: (device_names.size() > 0): 0!>0 No device specified.
• 版本、环境信息：    1）Paddle Lite 版本：v2.11    2）Host 环境：瑞芯微M1808-EVB    3）运行设备环境：请描述运行设备信息，如小米 9、iPhone13
• 问题描述： 在瑞芯微M1808开发板上运行目标检测，提示Check failed: (device_names.size() > 0): 0!>0 No device specified.

[paddle-bot[bot]]

您好，我们已经收到了您的问题，会安排技术人员尽快解答您的问题，请耐心等待。请您再次检查是否提供了清晰的问题描述、复现代码、环境&版本、报错信息等。同时，您也可以通过查看官网文档、常见问题、历史Issue来寻求解答。祝您生活愉快～

Hi! We've received your issue and please be patient to get responded. We will arrange technicians to answer your questions as soon as possible. Please make sure that you have posted enough message to demo your request. You may also check out the API，FAQ and Github Issue to get the answer.Have a nice day!

[yingshengBD]

你好 ，请先参考这个文档部署1808，https://paddle-lite.readthedocs.io/zh/develop/demo_guides/verisilicon_timvx.html 注意文档中描述的驱动版本

[ysjhlnu]

你好 ，请先参考这个文档部署1808，https://paddle-lite.readthedocs.io/zh/develop/demo_guides/verisilicon_timvx.html 注意文档中描述的驱动版本

驱动版本：6.4.3.5.293908 驱动和依赖都没问题 使用示例程序和模型不出现(device_names.size() > 0): 0!>0 No device specified这个问题，更换了模型就会出现这个问题。

[yingshengBD]

你是使用opt工具生成的模型？注意target 得同时包含 verisilicon_timvx和arm

[ysjhlnu]

你是使用opt工具生成的模型？注意target 得同时包含 verisilicon_timvx和arm

--valid_targets=rockchip_npu,arm --valid_targets=verisilicon_timvx,arm 用opt工具生成的模型，这两种类型都试过都提示这个问题

[hong19860320]

最好把原始的 paddle 模型放上来，方便我们复现哈~

[ysjhlnu]

最好把原始的 paddle 模型放上来，方便我们复现哈~

模型文件： 链接：https://pan.baidu.com/s/1CdvNpdYhQU56HeNJA19vmw 提取码：6v0c

配置文件使用的是yolov3_mobilenet_v1

[ysjhlnu]

使用官方示例程序也出现这个问题，详细执行过程如下: 194 RKNPU GraphImpl(46): [5 3/28 6: 5:51.355 ...dapter/nnadapter/src/utility/modeling.cc:937 SortOperationsInTopologicalOrder] model total operands: 496 [5 3/28 6: 5:51.355 ...dapter/nnadapter/src/utility/modeling.cc:937 SortOperationsInTopologicalOrder] model input operands: 1 [5 3/28 6: 5:51.355 ...dapter/nnadapter/src/utility/modeling.cc:937 SortOperationsInTopologicalOrder] model output operands: 2 [5 3/28 6: 5:51.356 ...dapter/nnadapter/src/utility/modeling.cc:937 SortOperationsInTopologicalOrder] model total operations: 77 [5 3/28 6: 5:51.356 ...iver/rockchip_npu/converter/converter.cc:39 Apply] Converting CONV_2D ... [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] input: 0x1F80AB70 precision: FLOAT32 layout: NCHW lifetime: MODEL_INPUT dimensions: [1,3,300,300,]

[5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] group = 1 [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] depthwise mode(0). [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] filter: 0x1F81E5E0 precision: FLOAT32 layout: NCHW lifetime: CONSTANT_REFERENCE dimensions: [32,3,3,3,]

[5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] input_channel_size: 3 [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] output_channel_size: 32 [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] filter_channel_size: 3 [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] filter_height: 3 [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] filter_width: 3 [5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] bias: 0x1F7D95F0 precision: FLOAT32 layout: NCHW lifetime: CONSTANT_REFERENCE dimensions: [32,]

[5 3/28 6: 5:51.356 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] auto_pad: AUTO_PAD_NONE [5 3/28 6: 5:51.357 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] paddings = [1, 1, 1, 1] [5 3/28 6: 5:51.357 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] strides = [2, 2] [5 3/28 6: 5:51.357 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] dilations = [1, 1] [5 3/28 6: 5:51.357 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] fuse_code = 1 [5 3/28 6: 5:51.357 .../driver/rockchip_npu/converter/conv2d.cc:24 ConvertConv2D] output: 0x1F811820 precision: FLOAT32 layout: NCHW lifetime: TEMPORARY_VARIABLE dimensions: [1,32,150,150,]

[F 3/28 6: 5:51.357 ...apter/src/driver/rockchip_npu/utility.cc:177 CreateRknnTensor] Can not add a rk::nn::Tensor with precision=FLOAT32 ! [F 3/28 6: 5:51.357 ...apter/src/driver/rockchip_npu/utility.cc:177 CreateRknnTensor] Can not add a rk::nn::Tensor with precision=FLOAT32 !

[F 3/28 6: 5:51.358 ...ter/nnadapter/src/runtime/compilation.cc:98 ~Program] Check failed: device_context: No device found. [F 3/28 6: 5:51.358 ...ter/nnadapter/src/runtime/compilation.cc:98 ~Program] Check failed: device_context: No device found.

terminate called after throwing an instance of 'nnadapter::logging::Exception' what(): NNAdapter C++ Exception: [F 3/28 6: 5:51.358 ...ter/nnadapter/src/runtime/compilation.cc:98 ~Program] Check failed: device_context: No device found.

./run.sh: line 114: 1928 Aborted (core dumped) ./$BUILD_DIR/ssd_detection_demo ../assets/models/$MODEL_NAME $MODEL_TYPE ../assets/labels/$LABEL_NAME ../assets/images/$IMAGE_NAME ../assets/results/$RESULT_NAME $NNADAPTER_DEVICE_NAMES $NNADAPTER_CONTEXT_PROPERTIES $NNADAPTER_MODEL_CACHE_DIR $NNADAPTER_MODEL_CACHE_TOKEN $NNADAPTER_SUBGRAPH_PARTITION_CONFIG_PATH

[yingshengBD]

是这样的，我们是建议使用 芯原TIMVX 既https://paddle-lite.readthedocs.io/zh/develop/demo_guides/verisilicon_timvx.html 这个方式来部署rk1808，这个要求驱动就是6.4.3.5。你的已经符合。 但是看你的日志，你是用的rk ddk来部署，那个就是要求驱动6.4.0，这个在rk ddk的文档中也有明确说明。 建议按照 https://paddle-lite.readthedocs.io/zh/develop/demo_guides/verisilicon_timvx.html 这个方法