search

PaddlePaddle / Paddle-Lite

PaddlePaddle High Performance Deep Learning Inference Engine for Mobile and Edge (飞桨高性能深度学习端侧推理引擎）
https://www.paddlepaddle.org.cn/lite
Apache License 2.0
6.94k stars 1.61k forks source link

Check failed: (idx < BlocksSize()): 0!<0 idx >= blocks.size() Aborted (core dumped) #4717

Closed maunzec closed 3 years ago

maunzec commented 3 years ago

python D:/Work/Paddle/PaddleDetection/tools/export_model.py -c ppyolo_mobilenet_v3_large.yml -o weights=output/ppyolo_mobilenet_v3_large/model_final --output_dir=./inference_model D:\Work\Paddle\PaddleDetection\ppdet\core\workspace.py:118: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated since Python 3.3,and in 3.9 it will stop working isinstance(merge_dct[k], collections.Mapping)): D:\Python37\lib\site-packages\paddle\fluid\layers\math_op_patch.py:278: UserWarning: D:\Work\Paddle\PaddleDetection\ppdet\modeling\backbones\mobilenet_v3.py:228 The behavior of expression A B has been unified with elementwise_mul(X, Y, axis=-1) from Paddle 2.0. If your code works well in the older versions but crashes in this version, try to use elementwise_mul(X, Y, axis=0) instead of A B. This transitional warning will be dropped in the future. op_type, op_type, EXPRESSION_MAP[method_name])) 2020-11-11 10:49:46,771-INFO: Load categories from D:/Work/CYY_glj/data/glj-number/label_list.txt 2020-11-11 10:49:46,774-INFO: Export inference config file to ./inference_model\ppyolo_mobilenet_v3_large\infer_cfg.yml 2020-11-11 10:49:46,881-INFO: save_inference_model pruned unused feed variables im_id 2020-11-11 10:49:46,882-INFO: Export inference model to ./inference_model\ppyolo_mobilenet_v3_large, input: ['image', 'im_size'], output: ['multiclass_nms_0.tmp_0']...

在arm上用编译好的paddllite,

root@orangepi3:~/work/ppyolo_mobilenet_v3_large# paddle_lite_opt --param_file=model --model_file=params --valid_targets=arm --optimize_out_type=naive_buffer --optimize_out=model [I 11/11 4: 1:59.263 ...oot/paddle-lite/lite/core/device_info.cc:1065 Setup] ARM multiprocessors name: [I 11/11 4: 1:59.263 ...oot/paddle-lite/lite/core/device_info.cc:1066 Setup] ARM multiprocessors number: 4 [I 11/11 4: 1:59.263 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 0, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 4: 1:59.263 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 1, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 4: 1:59.263 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 2, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 3, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1074 Setup] L1 DataCache size is: [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1078 Setup] L2 Cache size is: [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1082 Setup] L3 Cache size is: [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 4: 1:59.264 ...oot/paddle-lite/lite/core/device_info.cc:1086 Setup] Total memory: 2027832KB [I 11/11 4: 1:59.319 /root/paddle-lite/lite/api/cxx_api.cc Build:262] Load model from file. [F 11/11 4: 1:59.361 ...ite/model_parser/general/program_desc.cc:29 GetBlockpaddle::lite::general::BlockDesc] Check failed: (idx < BlocksSize()): 0!<0 idx >= blocks.size() Aborted (core dumped)

https://github.com/PaddlePaddle/PaddleDetection/issues/1689PaddleDetection 也提问了

maunzec commented 3 years ago

Paddle有要求？lite 2.7 2.6我都试过了，都不行，Paddle用过1.72，1.8.5，1.8.4，2.0

ysh329 commented 3 years ago

你好，你用的是什么模型呢，这个排版有点乱，lite的opt接收Paddle的模型转换，使用方式见：https://paddle-lite.readthedocs.io/zh/latest/user_guides/model_optimize_tool.html

感觉是你的模型是不是格式有问题，要求格式是model和params这种combined，或者是model和一堆拆散的权重文件这种non_combined形式。

maunzec commented 3 years ago

你好，你用的是什么模型呢，这个排版有点乱，lite的opt接收Paddle的模型转换，使用方式见：https://paddle-lite.readthedocs.io/zh/latest/user_guides/model_optimize_tool.html

感觉是你的模型是不是格式有问题，要求格式是model和params这种combined，或者是model和一堆拆散的权重文件这种non_combined形式。

从连接上下载的

export_model.py -c ppyolo.yml -o weights=ppyolo --output_dir=./inference_model

root@orangepi3:~/work/ppyolo# paddle_lite_opt --param_file=params --model_file=model --valid_targets=arm --optimize_out_type=naive_buffer --optimize_out=model

root@mazec-ubuntu:/lite# cd ppyolo root@mazec-ubuntu:/lite/ppyolo# /root/lite/model_optimize_tool --param_file=params--model_file=model --valid_targets=arm --optimize_out_type=naive_buffer --optimize_out=model

maunzec commented 3 years ago

无论是自己训练的，还是官网下载的，都不可以

maunzec commented 3 years ago

https://github.com/PaddlePaddle/PaddleDetection/issues/1689，，这个是我之前提问的

ysh329 commented 3 years ago

比方这种方式：

./opt --model_dir=./mobilenet_v1_quant \
      --optimize_out_type=naive_buffer \
      --optimize_out=mobilenet_v1_quant_opt \
      --valid_targets=arm

这种是针对cmobined模型的

ysh329 commented 3 years ago

我看你的模型后缀是yml格式，应该是model（结构文件），和一个param（参数文件，或者是拆散的很多个权重参数文件）。

但无论怎么说，都应该还有一个model文件，不应该是yml

maunzec commented 3 years ago

排版有问题，我等一会截图把

maunzec commented 3 years ago

我看你的模型后缀是yml格式，应该是model（结构文件），和一个param（参数文件，或者是拆散的很多个权重参数文件）。

但无论怎么说，都应该还有一个model文件，不应该是yml

那个用paddle 导出模型

paddle_lite_opt --param_file=params --model_file=model --valid_targets=arm --optimize_out_type=naive_buffer --optimize_out=model 这是用编译后lite 2.7 转换模型

[I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1065 Setup] ARM multiprocessors name: [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1066 Setup] ARM multiprocessors number: 4 [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 0, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 1, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 2, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 3, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1074 Setup] L1 DataCache size is: [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1078 Setup] L2 Cache size is: [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1082 Setup] L3 Cache size is: [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/11 14: 4:56.221 ...oot/paddle-lite/lite/core/device_info.cc:1086 Setup] Total memory: 2027832KB [I 11/11 14: 4:56.340 /root/paddle-lite/lite/api/cxx_api.cc Build:262] Load model from file. [F 11/11 14: 4:56.665 /root/paddle-lite/lite/core/program.cc Build:326] Check failed: op: no Op found for hard_sigmoid Aborted (core dumped)

[F 11/11 14: 4:56.665 /root/paddle-lite/lite/core/program.cc Build:326] Check failed: op: no Op found for hard_sigmoid Aborted (core dumped) 这个是什么问题

maunzec commented 3 years ago

图片

maunzec commented 3 years ago

我看你的模型后缀是yml格式，应该是model（结构文件），和一个param（参数文件，或者是拆散的很多个权重参数文件）。

但无论怎么说，都应该还有一个model文件，不应该是yml

[I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1065 Setup] ARM multiprocessors name: [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1066 Setup] ARM multiprocessors number: 4 [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 0, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 1, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 2, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 3, max freq: 1488, min freq: 1488, cluster ID: 0, CPU ARCH: A53 [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1074 Setup] L1 DataCache size is: [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1076 Setup] 32 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1078 Setup] L2 Cache size is: [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1080 Setup] 512 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1082 Setup] L3 Cache size is: [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1084 Setup] 0 KB [I 11/12 1:49:28.427 ...oot/paddle-lite/lite/core/device_info.cc:1086 Setup] Total memory: 2027832KB [W 11/12 1:49:28.428 ...e-lite/lite/model_parser/model_parser.cc:954 LoadModelFbsFromFile] warning: the version of opt that transformed this model is not consistent with current Paddle-Lite version. version of opt:2a2e6fab version of current Paddle-Lite:2a2e6fa [F 11/12 1:49:28.548 /root/paddle-lite/lite/core/program.cc RuntimeProgram:187] Check failed: op: no Op found for hard_sigmoid Aborted (core dumped)

模型在x86 ubuntu上转换的，在arm上运行，报错

ysh329 commented 3 years ago

那应该是缺少这个Op，你用的是什么版本呀，2.7还是2,7beta呀

maunzec commented 3 years ago

2.7 2a2e6fab

maunzec commented 3 years ago

Op

Op是什么组件

ysh329 commented 3 years ago

Op

Op是什么组件

就是缺少这么个算子，但这个hard_sigmoid理论上应该是有arm cpu实现的

ysh329 commented 3 years ago

我看看你的模型文件有哪些，你ls一下（显示的model等等那些文件）

maunzec commented 3 years ago

图片 ppyolo_mobilenet_v3_large 这个训练的

ysh329 commented 3 years ago

你转nb格式的模型成功了吗，paddle-lite只能跑nb格式的模型（就是用opt转换后的）

maunzec commented 3 years ago

成功的

maunzec commented 3 years ago

ls PPopt.py model params infer_cfg.yml root@orangepi3:~/work/ppyolo_mobilenet_v3_large#

ysh329 commented 3 years ago

OK 我看到了，有点奇怪，你能把这个Fluid模型发给我一份嘛，或者把地址留一下，我看下

ysh329 commented 3 years ago

Fluid模型就是model，以及对应权重文件的格式

maunzec commented 3 years ago

ppyolo_mobilenet_v3_large.zip

maunzec commented 3 years ago

github 上传的你能下载吗？或者我传到奶牛快传，网盘

ysh329 commented 3 years ago

可以下载的

ysh329 commented 3 years ago

你这个opt是自己编译的嘛？

maunzec commented 3 years ago

你这个opt是自己编译的嘛？

是的

warning: the version of opt that transformed this model is not consistent with current Paddl version of opt:2a2e6fab version of current Paddle-Lite:2a2e6fa [F 11/12 1:50:58.337 /root/paddle-lite/lite/core/program.cc RuntimeProgram:187] Check failed: op: no Op found for hard_sigmoid

ysh329 commented 3 years ago

哦哦好的，我切到这个commit，不过有个问题，还是使用我们发布的版本吧，就是release页面的

ysh329 commented 3 years ago

要不这样吧，试试https://github.com/PaddlePaddle/Paddle-Lite/releases，这个beta1，起码这里的是我们有较全面测试的

ysh329 commented 3 years ago

我试试beta1

maunzec commented 3 years ago

git clone -b release/v2.7 https://gitee.com/paddlepaddle/paddle-lite.git 我是在gitee上获取的

ysh329 commented 3 years ago

opt版本和库版本最好是要一样

maunzec commented 3 years ago

ok，我试一试

ysh329 commented 3 years ago

我也试试beta1

ysh329 commented 3 years ago

我这边试了下2.7beta1转换模型是OK的，正编译库中

ysh329 commented 3 years ago

我这边beta1的库也是OK的，日志如下：

$ ./run.sh
++ ADB='adb -s 4c4f947c '
++ ADB='adb -s 328fca91 '
++ rm mobilenetv1_light_api
++ adb -s 328fca91 shell rm /data/local/tmp/lite-ppyolo-opt/mobilenetv1_light_api
++ wget -c http://10.87.145.34:8889/code/lite-ppyolo-opt/build.lite.android.armv7.clang.opencl/inference_lite_lib.android.armv7.opencl/demo/cxx/mobile_light/mobilenetv1_light_api
--2020-11-12 11:25:31--  http://10.87.145.34:8889/code/lite-ppyolo-opt/build.lite.android.armv7.clang.opencl/inference_lite_lib.android.armv7.opencl/demo/cxx/mobile_light/mobilenetv1_light_api
Connecting to 10.87.145.34:8889... connected.
HTTP request sent, awaiting response... 200 OK
Length: 9061736 (8.6M) [application/octet-stream]
Saving to: ‘mobilenetv1_light_api’

mobilenetv1_light_api                             100%[============================================================================================================>]   8.64M  6.95MB/s    in 1.2s

2020-11-12 11:25:32 (6.95 MB/s) - ‘mobilenetv1_light_api’ saved [9061736/9061736]

++ wget -c http://10.87.145.34:8889/code/lite-ppyolo-opt/ppyolo_mobilenetv3_large_arm_beta1.nb
--2020-11-12 11:25:32--  http://10.87.145.34:8889/code/lite-ppyolo-opt/ppyolo_mobilenetv3_large_arm_beta1.nb
Connecting to 10.87.145.34:8889... connected.
HTTP request sent, awaiting response... 200 OK

    The file is already fully retrieved; nothing to do.

++ adb -s 328fca91 shell mkdir /data/local/tmp/lite-ppyolo-opt/
mkdir: '/data/local/tmp/lite-ppyolo-opt/': File exists
++ adb -s 328fca91 shell rm /data/local/tmp/lite-ppyolo-opt/mobilenetv1_light_api
rm: /data/local/tmp/lite-ppyolo-opt/mobilenetv1_light_api: No such file or directory
++ adb -s 328fca91 push ppyolo_mobilenetv3_large_arm_beta1.nb /data/local/tmp/lite-ppyolo-opt/ppyolo_mobilenetv3_large_arm_beta1.nb
2600 KB/s (18712194 bytes in 7.027s)
++ adb -s 328fca91 push mobilenetv1_light_api /data/local/tmp/lite-ppyolo-opt/mobilenetv1_light_api
1891 KB/s (9061736 bytes in 4.677s)
++ adb -s 328fca91 shell chmod +x /data/local/tmp/lite-ppyolo-opt/mobilenetv1_light_api
++ adb -s 328fca91 shell 'export GLOG_v=0; /data/local/tmp/lite-ppyolo-opt/mobilenetv1_light_api /data/local/tmp/lite-ppyolo-opt/ppyolo_mobilenetv3_large_arm_beta1.nb 1,3,608,608:1,2 100 1 0'
raw_input_shapes: 1,3,608,608:1,2
input shape: 1,3,608,608
input shape: 1,2
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1065 Setup] ARM multiprocessors name: MODEL NAME  : ARMV8 PROCESSOR REV 4 (V8L)
HARDWARE    : QUALCOMM TECHNOLOGIES, INC MSM8998
_MSM8998_MSM8998_
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1066 Setup] ARM multiprocessors number: 8
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 0, max freq: 1900, min freq: 1900, cluster ID: 1, CPU ARCH: A53
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 1, max freq: 1900, min freq: 1900, cluster ID: 1, CPU ARCH: A53
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 2, max freq: 1900, min freq: 1900, cluster ID: 1, CPU ARCH: A53
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 3, max freq: 1900, min freq: 1900, cluster ID: 1, CPU ARCH: A53
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 4, max freq: 2457, min freq: 2457, cluster ID: 0, CPU ARCH: A73
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 5, max freq: 2457, min freq: 2457, cluster ID: 0, CPU ARCH: A73
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 6, max freq: 2457, min freq: 2457, cluster ID: 0, CPU ARCH: A73
[I 11/12 11:25:55.544 ...lite-ppyolo-opt/lite/core/device_info.cc:1068 Setup] ARM multiprocessors ID: 7, max freq: 2457, min freq: 2457, cluster ID: 0, CPU ARCH: A73
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1074 Setup] L1 DataCache size is:
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 32 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 32 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 32 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 32 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 64 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 64 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 64 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1076 Setup] 64 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1078 Setup] L2 Cache size is:
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1080 Setup] 1024 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1082 Setup] L3 Cache size is:
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1084 Setup] 0 KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:1086 Setup] Total memory: 5862104KB
[I 11/12 11:25:55.545 ...lite-ppyolo-opt/lite/core/device_info.cc:439 check_cpu_online] CPU id:7 is offine
[W 11/12 11:25:55.546 ...lite-ppyolo-opt/lite/core/device_info.cc:1139 SetRunMode] Some cores are offline, switch to NO BIND MODE
[I 11/12 11:25:55.651 ...o-opt/lite/backends/opencl/cl_runtime.cc:49 Init] is_cl_runtime_initialized_:0
[I 11/12 11:25:55.651 ...o-opt/lite/backends/opencl/cl_runtime.cc:57 Init] opencl_lib_found:1
[I 11/12 11:25:55.651 ...o-opt/lite/backends/opencl/cl_runtime.cc:65 Init] dlsym_success:1
[I 11/12 11:25:55.655 ...o-opt/lite/backends/opencl/cl_runtime.cc:73 Init] is_platform_init:1
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:229 InitializeDevice] Using device: QUALCOMM Adreno(TM)
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:254 InitializeDevice] device_version:OpenCL 2.0 Adreno(TM) 540
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:256 InitializeDevice] device_type:GPU
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:260 InitializeDevice] The chosen device has 4 compute units.
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:264 InitializeDevice] CL_DEVICE_MAX_CLOCK_FREQUENCY:1
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:274 InitializeDevice] The local memory size of the chosen device is 32.000000 KB.
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:281 InitializeDevice] CL_DEVICE_GLOBAL_MEM_CACHE_SIZE(KB):128.000000 KB.
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:289 InitializeDevice] CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE(KB):0.062500 KB.
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:296 InitializeDevice] CL_DEVICE_GLOBAL_MEM_SIZE(KB):2931052.000000 KB.
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:304 InitializeDevice] CL_DEVICE_MAX_WORK_GROUP_SIZE:1024
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:308 InitializeDevice] CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS:3
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:313 InitializeDevice] max_work_item_sizes[0]:1024
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:313 InitializeDevice] max_work_item_sizes[1]:1024
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:313 InitializeDevice] max_work_item_sizes[2]:1024
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:324 InitializeDevice] CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE:64.000000
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:335 InitializeDevice] The chosen device supports image processing.
[I 11/12 11:25:55.656 ...o-opt/lite/backends/opencl/cl_runtime.cc:344 InitializeDevice] CL_DEVICE_IMAGE2D_MAX_HEIGHT:16384
[I 11/12 11:25:55.657 ...o-opt/lite/backends/opencl/cl_runtime.cc:348 InitializeDevice] CL_DEVICE_IMAGE2D_MAX_WIDTH:16384
[I 11/12 11:25:55.657 ...o-opt/lite/backends/opencl/cl_runtime.cc:357 InitializeDevice] The chosen device supports the half data type.
[I 11/12 11:25:55.657 ...o-opt/lite/backends/opencl/cl_runtime.cc:365 InitializeDevice] CL_DEVICE_ADDRESS_BITS:32
[I 11/12 11:25:55.662 ...o-opt/lite/backends/opencl/cl_runtime.cc:369 InitializeDevice] CL_DRIVER_VERSION:OpenCL 2.0 QUALCOMM build: commit #9c9b012 changeid #I92eb381bc9 Date: 12/31/17 Sun Local Branch:  Remote Branch: refs/tags/AU_LINUX_ANDROID_LA.UM.6.4.R1.08.00.00.309.049 Compiler E031.35.02.02
[I 11/12 11:25:55.663 ...o-opt/lite/backends/opencl/cl_runtime.cc:81 Init] is_device_init:1
[I 11/12 11:25:55.663 ...o-opt/lite/backends/opencl/cl_runtime.cc:399 GetAdrenoContextProperties] GPUPerfMode::PERF_HIGH
[I 11/12 11:25:55.663 ...o-opt/lite/backends/opencl/cl_runtime.cc:418 GetAdrenoContextProperties] GPUPriorityLevel::PRIORITY_HIGH
[I 11/12 11:25:55.667 ...o-opt/lite/backends/opencl/cl_runtime.cc:93 Init] set is_cl_runtime_initialized_ = true
[I 11/12 11:25:55.667 ...o-opt/lite/backends/opencl/cl_runtime.cc:49 Init] is_cl_runtime_initialized_:1
[I 11/12 11:25:55.667 ...lo-opt/lite/backends/opencl/cl_runtime.h:80 OpenCLAvaliableForDevice] support_fp16:1
[I 11/12 11:25:55.667 ...lo-opt/lite/backends/opencl/cl_runtime.h:86 OpenCLAvaliableForDevice] is_device_avaliable_for_opencl_:1
[I 11/12 11:25:55.667 ...o-opt/lite/backends/opencl/cl_runtime.cc:49 Init] is_cl_runtime_initialized_:1
[I 11/12 11:25:55.667 ...o-opt/lite/backends/opencl/cl_runtime.cc:49 Init] is_cl_runtime_initialized_:1
[I 11/12 11:25:55.672 ...o-opt/lite/backends/opencl/cl_runtime.cc:49 Init] is_cl_runtime_initialized_:1
[I 11/12 11:25:55.672 ...lo-opt/lite/backends/opencl/cl_context.h:42 ~CLContext] release cl::Program, cl::Kernel finished.
input_shapes.size():2
run_idx:1 / 100: 530.094 ms
run_idx:2 / 100: 531.469 ms
run_idx:3 / 100: 532.707 ms
run_idx:4 / 100: 532.015 ms
run_idx:5 / 100: 529.358 ms
run_idx:6 / 100: 532.445 ms
run_idx:7 / 100: 531.26 ms
run_idx:8 / 100: 531.12 ms
run_idx:9 / 100: 530.802 ms
run_idx:10 / 100: 530.433 ms
run_idx:11 / 100: 531.888 ms
run_idx:12 / 100: 532.118 ms
run_idx:13 / 100: 529.417 ms
run_idx:14 / 100: 531.428 ms
run_idx:15 / 100: 531.876 ms
run_idx:16 / 100: 531.482 ms
run_idx:17 / 100: 530.298 ms
run_idx:18 / 100: 531.065 ms
run_idx:19 / 100: 531.348 ms
run_idx:20 / 100: 531.671 ms
run_idx:21 / 100: 531.629 ms
run_idx:22 / 100: 529.569 ms
run_idx:23 / 100: 532.523 ms
run_idx:24 / 100: 534.141 ms
run_idx:25 / 100: 532.028 ms
run_idx:26 / 100: 529.419 ms
run_idx:27 / 100: 532.084 ms
run_idx:28 / 100: 529.471 ms
run_idx:29 / 100: 531.707 ms
run_idx:30 / 100: 531.552 ms
run_idx:31 / 100: 530.931 ms
run_idx:32 / 100: 531.393 ms
run_idx:33 / 100: 531.922 ms
run_idx:34 / 100: 532.595 ms
run_idx:35 / 100: 531.487 ms
run_idx:36 / 100: 530.623 ms
run_idx:37 / 100: 531.94 ms
run_idx:38 / 100: 531.071 ms
run_idx:39 / 100: 530.267 ms
run_idx:40 / 100: 530.282 ms
run_idx:41 / 100: 529.852 ms
run_idx:42 / 100: 531.245 ms
run_idx:43 / 100: 532.453 ms
run_idx:44 / 100: 529.838 ms
run_idx:45 / 100: 529.822 ms
run_idx:46 / 100: 530.146 ms
run_idx:47 / 100: 532.568 ms
run_idx:48 / 100: 530.402 ms
run_idx:49 / 100: 520.666 ms
run_idx:50 / 100: 516.877 ms
run_idx:51 / 100: 514.35 ms
run_idx:52 / 100: 515.965 ms
run_idx:53 / 100: 514.164 ms
run_idx:54 / 100: 515.38 ms
run_idx:55 / 100: 514.994 ms
run_idx:56 / 100: 514.632 ms
run_idx:57 / 100: 517.226 ms
run_idx:58 / 100: 514.249 ms
run_idx:59 / 100: 516.129 ms
run_idx:60 / 100: 522.263 ms
run_idx:61 / 100: 524.967 ms
run_idx:62 / 100: 556.351 ms
run_idx:63 / 100: 562.927 ms
run_idx:64 / 100: 555.312 ms
run_idx:65 / 100: 553.707 ms
run_idx:66 / 100: 551.012 ms
run_idx:67 / 100: 552.859 ms
run_idx:68 / 100: 551.247 ms
run_idx:69 / 100: 553.683 ms
run_idx:70 / 100: 553.039 ms
run_idx:71 / 100: 554.428 ms
run_idx:72 / 100: 558.777 ms
run_idx:73 / 100: 557.634 ms
run_idx:74 / 100: 555.016 ms
run_idx:75 / 100: 549.018 ms
run_idx:76 / 100: 552.123 ms
run_idx:77 / 100: 548.147 ms
run_idx:78 / 100: 552.653 ms
run_idx:79 / 100: 551.36 ms
run_idx:80 / 100: 555.862 ms
run_idx:81 / 100: 552.851 ms
run_idx:82 / 100: 553.772 ms
run_idx:83 / 100: 555.917 ms
run_idx:84 / 100: 552.496 ms
run_idx:85 / 100: 551.873 ms
run_idx:86 / 100: 550.135 ms
run_idx:87 / 100: 551.849 ms
run_idx:88 / 100: 553.691 ms
run_idx:89 / 100: 550.171 ms
run_idx:90 / 100: 552.302 ms
run_idx:91 / 100: 551.093 ms
run_idx:92 / 100: 551.774 ms
run_idx:93 / 100: 556.085 ms
run_idx:94 / 100: 554.689 ms
run_idx:95 / 100: 554.505 ms
run_idx:96 / 100: 555.384 ms
run_idx:97 / 100: 553.183 ms
run_idx:98 / 100: 552.761 ms
run_idx:99 / 100: 548.615 ms
run_idx:100 / 100: 548.308 ms

======= benchmark summary =======
input_shape(s) (NCHW):1,3,608,608, : 1,2,
model_dir:/data/local/tmp/lite-ppyolo-opt/ppyolo_mobilenetv3_large_arm_beta1.nb
warmup:1
repeats:100
max_duration:562.927
min_duration:514.164
avg_duration:537.957

====== output summary ======
output tensor num:1

--- output tensor 0 ---
output shape(NCHW):6 6
output tensor 0 elem num:36
output tensor 0 standard deviation:4.97388e+08
output tensor 0 mean value:3.80072e+08
[I 11/12 11:26:50.391 ...o-opt/lite/backends/opencl/cl_runtime.cc:29 ~CLRuntime] is_cl_runtime_initialized_:1
ysh329 commented 3 years ago

lite-ppyolo-arm-cpu-unit-test-beta1.zip

ysh329 commented 3 years ago

我上传了我转换的模型，以及验证用的单元测试，就是上面这个压缩包里有个bin文件，一个run.sh的脚本，一个转好的cpu模型，你也可以跑跑看，可能需要修下run.sh里的机器码为自己插电脑的手机serial，就可以验证了

maunzec commented 3 years ago

print_supported_ops.txt 编译了2.7 beta1，结果也是说不支持op

[F 11/12 7:43:38.921 /root/paddle-lite/lite/core/program.cc RuntimeProgram:187] Check failed: op: no Op found for hard_sigmoid Aborted (core dumped)

maunzec commented 3 years ago

print_supported_ops.txt 里面有paddle_lite_opt --print_supported_ops=true --valid_targets=arm 输出的结果 print_supported_ops.txt

ysh329 commented 3 years ago

我上面提供的脚本可以跑么

maunzec commented 3 years ago

不行啊，我这个是armv8的开发板

maunzec commented 3 years ago

Ubuntu 18.04.4 LTS 全志H6 的芯片

ysh329 commented 3 years ago

是安卓系统不？是arm cpu吧，那就能跑的， 我这边能在安卓手机上跑的，安卓手机是armv7/v8的

maunzec commented 3 years ago

root@orangepi3:~/work/lite-ppyolo# ls 2020_0000022.jpg lite.py mobilenetv1_light_api mobilenetv1_light_api.py ppyolo_mobilenetv3_large_arm_beta1.nb run.sh root@orangepi3:~/work/lite-ppyolo# ./mobilenetv1_light_api ppyolo_mobilenetv3_large_arm_beta1.nb -bash: ./mobilenetv1_light_api: No such file or directory root@orangepi3:~/work/lite-ppyolo# chmod +x mobilenetv1_light_api root@orangepi3:~/work/lite-ppyolo# ./mobilenetv1_light_api ppyolo_mobilenetv3_large_arm_beta1.nb -bash: ./mobilenetv1_light_api: No such file or directory root@orangepi3:~/work/lite-ppyolo# mobilenetv1_light_api -bash: mobilenetv1_light_api: command not found root@orangepi3:~/work/lite-ppyolo# mobilenetv1_light_api ppyolo_mobilenetv3_large_arm_beta1.nb -bash: mobilenetv1_light_api: command not found root@orangepi3:~/work/lite-ppyolo# mobilenetv1_light_api ppyolo_mobilenetv3_large_arm_beta1.nb 1,3,608,608:1,2 100 1 0 -bash: mobilenetv1_light_api: command not found

maunzec commented 3 years ago

图片

maunzec commented 3 years ago

给权限了也无法执行

ysh329 commented 3 years ago

你这个是arm linux吧？不是android吧？那就不能这么执行，我上面是针对安卓的，老哥啊

maunzec commented 3 years ago

是的啊

ysh329 commented 3 years ago

你要编译针对arm linux的库