安装和初步使用 nn-Meter

已于 2023-11-21 08:59:31 修改
阅读量484 收藏 1
点赞数 1
分类专栏： 20240202 文章标签： 安装和初步使用 nn-Meter
于 2023-11-20 23:16:26 首次发布
世上没有白读的书，每一页都算数。
本文链接：https://blog.csdn.net/chengyq116/article/details/134497336
版权

安装和初步使用 nn-Meter

nn-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices
nn-Meter：精准预测深度学习模型在边缘设备上的推理延迟

Li Lyna Zhang, Shihao Han, Jianyu Wei, Ningxin Zheng, Ting Cao, Yuqing Yang, Yunxin Liu

1. nn-Meter

https://github.com/microsoft/nn-Meter

nn-Meter is a novel and efficient system to accurately predict the inference latency of DNN models on diverse edge devices. The key idea is dividing a whole model inference into kernels, i.e., the execution units of fused operators on a device, and conduct kernel-level prediction. We currently evaluate four popular platforms on a large dataset of 26k models. It achieves 99.0% (mobile CPU), 99.1% (mobile Adreno 640 GPU), 99.0% (mobile Adreno 630 GPU), and 83.4% (Intel VPU) prediction accuracy.
为了高效、准确地预测深度神经网络模型在不同边缘设备上的推理延迟，作者提出并开发了一个基于内核的模型推理延迟预测系统 nn-Meter，引入了内核检测，可找出算子融合行为。通过对最有价值的数据进行采样，nn-Meter 有效地建立了内核的延迟预测器。

The current supported hardware and inference frameworks:

DeviceFrameworkProcessor±10% AccuracyHardware name
Pixel4TFLite v2.1CortexA76 CPU99.0%cortexA76cpu_tflite21
Mi9TFLite v2.1Adreno 640 GPU99.1%adreno640gpu_tflite21
Pixel3XLTFLite v2.1Adreno 630 GPU99.0%adreno630gpu_tflite21
Intel Movidius NCS2OpenVINO2019R2Myriad VPU83.4%myriadvpu_openvino2019r2

Tags
https://github.com/microsoft/nn-Meter/tags

Releases
https://github.com/microsoft/nn-Meter/releases

https://github.com/microsoft/nn-Meter/releases/tag/v1.0-data

adreno630gpu_tflite21.zip
adreno640gpu_tflite21.zip
cortexA76cpu_tflite21.zip
datasets.zip
ir_graphs.zip
myriadvpu_openvino2019r2.zip
onnx_models.zip
pb_models.zip
Source code (zip)
Source code (tar.gz)

https://github.com/microsoft/nn-Meter/releases/tag/v2.0-data

tflite_benchmark_tools_v2.1.zip
tflite_benchmark_tools_v2.7.zip
Source code (zip)
Source code (tar.gz)

nn-Meter Builder
https://github.com/microsoft/nn-Meter/blob/main/docs/builder/overview.md

Neural Network Intelligence，NNI
https://github.com/microsoft/nni
https://www.microsoft.com/en-us/research/project/neural-network-intelligence/
https://nni.readthedocs.io/zh/stable/
https://nni.readthedocs.io/en/stable/

NNI (Neural Network Intelligence) is a toolkit to help users run automated machine learning (AutoML) experiments.

26k latency benchmark dataset
https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/datasets.zip

2. Installation

Currently nn-Meter has been tested on Linux and Windows system. Windows 10, Ubuntu 16.04 and 20.04 with python 3.6.10 are tested and supported. Please first install python3 before nn-Meter installation. Then nn-Meter could be installed by running:

pip install nn-meter

pip install nn-meter --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple

nn-meter==2.0 has been released now.

If you want to try latest code, please install nn-Meter from source code. First git clone nn-Meter package to local:

git clone git@github.com:microsoft/nn-Meter.git
cd nn-Meter

Then simply run the following pip install in an environment that has python >= 3.6. The command will complete the automatic installation of all necessary dependencies and nn-Meter.

pip install .

nn-Meter is a latency predictor of models with type of Tensorflow, PyTorch, Onnx, nn-meter IR graph and NNI IR graph (https://github.com/microsoft/nni). To use nn-Meter for specific model type, you also need to install corresponding required packages. The well tested versions are listed below:

Testing Model TypeRequirements
Tensorflowtensorflow==2.6.0
Torchtorch==1.9.0, torchvision==0.10.0, (alternative) [onnx>=1.9.0, onnx-simplifier==0.3.6] or [nni>=2.4]
Onnxonnx==1.9.0
nn-Meter IR graph—
NNI IR graphnni>=2.4

Please also check the versions of numpy and scikit_learn. The different versions may change the prediction accuracy of kernel predictors.

The stable version of wheel binary package will be released soon.

2.1. python=3.7

conda create --name py37 python=3.7

(base) yongqiang@yongqiang:~$ conda env list
# conda environments:
#
base                  *  /home/yongqiang/miniconda3

(base) yongqiang@yongqiang:~$
(base) yongqiang@yongqiang:~$ conda create --name py37 python=3.7
Collecting package metadata (current_repodata.json): done
Solving environment: unsuccessful attempt using repodata from current_repodata.json, retrying with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: done

==> WARNING: A newer version of conda exists. <==
  current version: 23.7.3
  latest version: 23.10.0

Please update conda by running

    $ conda update -n base -c defaults conda

Or to minimize the number of packages updated during conda update use

     conda install conda=23.10.0

## Package Plan ##

  environment location: /home/yongqiang/miniconda3/envs/py37

  added / updated specs:
    - python=3.7

The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    ca-certificates-2023.08.22 |       h06a4308_0         123 KB  defaults
    certifi-2022.12.7          |   py37h06a4308_0         150 KB  defaults
    openssl-1.1.1w             |       h7f8727e_0         3.7 MB  defaults
    pip-22.3.1                 |   py37h06a4308_0         2.7 MB  defaults
    python-3.7.16              |       h7a1cb2a_0        44.8 MB  defaults
    setuptools-65.6.3          |   py37h06a4308_0         1.1 MB  defaults
    wheel-0.38.4               |   py37h06a4308_0          63 KB  defaults
    ------------------------------------------------------------
                                           Total:        52.7 MB

The following NEW packages will be INSTALLED:

  _libgcc_mutex      anaconda/pkgs/main/linux-64::_libgcc_mutex-0.1-main
  _openmp_mutex      anaconda/pkgs/main/linux-64::_openmp_mutex-5.1-1_gnu
  ca-certificates    anaconda/pkgs/main/linux-64::ca-certificates-2023.08.22-h06a4308_0
  certifi            anaconda/pkgs/main/linux-64::certifi-2022.12.7-py37h06a4308_0
  ld_impl_linux-64   anaconda/pkgs/main/linux-64::ld_impl_linux-64-2.38-h1181459_1
  libffi             anaconda/pkgs/main/linux-64::libffi-3.4.4-h6a678d5_0
  libgcc-ng          anaconda/pkgs/main/linux-64::libgcc-ng-11.2.0-h1234567_1
  libgomp            anaconda/pkgs/main/linux-64::libgomp-11.2.0-h1234567_1
  libstdcxx-ng       anaconda/pkgs/main/linux-64::libstdcxx-ng-11.2.0-h1234567_1
  ncurses            anaconda/pkgs/main/linux-64::ncurses-6.4-h6a678d5_0
  openssl            anaconda/pkgs/main/linux-64::openssl-1.1.1w-h7f8727e_0
  pip                anaconda/pkgs/main/linux-64::pip-22.3.1-py37h06a4308_0
  python             anaconda/pkgs/main/linux-64::python-3.7.16-h7a1cb2a_0
  readline           anaconda/pkgs/main/linux-64::readline-8.2-h5eee18b_0
  setuptools         anaconda/pkgs/main/linux-64::setuptools-65.6.3-py37h06a4308_0
  sqlite             anaconda/pkgs/main/linux-64::sqlite-3.41.2-h5eee18b_0
  tk                 anaconda/pkgs/main/linux-64::tk-8.6.12-h1ccaba5_0
  wheel              anaconda/pkgs/main/linux-64::wheel-0.38.4-py37h06a4308_0
  xz                 anaconda/pkgs/main/linux-64::xz-5.4.2-h5eee18b_0
  zlib               anaconda/pkgs/main/linux-64::zlib-1.2.13-h5eee18b_0

Proceed ([y]/n)? y

Downloading and Extracting Packages

Preparing transaction: done
Verifying transaction: done
Executing transaction: done
#
# To activate this environment, use
#
#     $ conda activate py37
#
# To deactivate an active environment, use
#
#     $ conda deactivate

(base) yongqiang@yongqiang:~$
(base) yongqiang@yongqiang:~$ conda env list
# conda environments:
#
base                  *  /home/yongqiang/miniconda3
py37                     /home/yongqiang/miniconda3/envs/py37

(base) yongqiang@yongqiang:~$
(base) yongqiang@yongqiang:~$ conda activate py37
(py37) yongqiang@yongqiang:~$ conda env list
# conda environments:
#
base                     /home/yongqiang/miniconda3
py37                  *  /home/yongqiang/miniconda3/envs/py37

(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ python
Python 3.7.16 (default, Jan 17 2023, 22:20:44)
[GCC 11.2.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> exit()
(py37) yongqiang@yongqiang:~$

2.2. nn-meter 2.0

pip install nn-meter --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple

(py37) yongqiang@yongqiang:~$ pip install nn-meter --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied: nn-meter in ./miniconda3/envs/py37/lib/python3.7/site-packages (2.0)
Requirement already satisfied: networkx in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (2.6.3)
Requirement already satisfied: scikit-learn in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (1.0.2)
Requirement already satisfied: numpy in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (1.21.6)
Requirement already satisfied: requests in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (2.31.0)
Requirement already satisfied: pandas in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (1.3.5)
Requirement already satisfied: jsonlines in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (3.1.0)
Requirement already satisfied: packaging in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (23.2)
Requirement already satisfied: PyYAML in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (6.0.1)
Requirement already satisfied: protobuf in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (4.24.4)
Requirement already satisfied: tqdm in ./miniconda3/envs/py37/lib/python3.7/site-packages (from nn-meter) (4.66.1)
Requirement already satisfied: attrs>=19.2.0 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from jsonlines->nn-meter) (23.1.0)
Requirement already satisfied: typing-extensions in ./miniconda3/envs/py37/lib/python3.7/site-packages (from jsonlines->nn-meter) (4.7.1)
Requirement already satisfied: python-dateutil>=2.7.3 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from pandas->nn-meter) (2.8.2)
Requirement already satisfied: pytz>=2017.3 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from pandas->nn-meter) (2023.3.post1)
Requirement already satisfied: certifi>=2017.4.17 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from requests->nn-meter) (2022.12.7)
Requirement already satisfied: idna<4,>=2.5 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from requests->nn-meter) (3.4)
Requirement already satisfied: urllib3<3,>=1.21.1 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from requests->nn-meter) (2.0.7)
Requirement already satisfied: charset-normalizer<4,>=2 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from requests->nn-meter) (3.3.2)
Requirement already satisfied: joblib>=0.11 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from scikit-learn->nn-meter) (1.3.2)
Requirement already satisfied: threadpoolctl>=2.0.0 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from scikit-learn->nn-meter) (3.1.0)
Requirement already satisfied: scipy>=1.1.0 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from scikit-learn->nn-meter) (1.7.3)
Requirement already satisfied: importlib-metadata in ./miniconda3/envs/py37/lib/python3.7/site-packages (from attrs>=19.2.0->jsonlines->nn-meter) (6.7.0)
Requirement already satisfied: six>=1.5 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from python-dateutil>=2.7.3->pandas->nn-meter) (1.16.0)
Requirement already satisfied: zipp>=0.5 in ./miniconda3/envs/py37/lib/python3.7/site-packages (from importlib-metadata->attrs>=19.2.0->jsonlines->nn-meter) (3.15.0)
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ conda list | grep scikit-learn
scikit-learn              1.0.2                    pypi_0    pypi
(py37) yongqiang@yongqiang:~$

(py37) yongqiang@yongqiang:~$ conda list | grep nn-meter
nn-meter                  2.0                      pypi_0    pypi
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ nn-meter -h
usage: nn-meter [-h] [-v] [--list-predictors] [--list-backends]
                [--list-kernels] [--list-operators] [--list-testcases]
                {predict,lat_pred,get_ir,create,connect,register,unregister}
                ...

please run "nn-meter {positional argument} --help" to see nn-meter guidance

positional arguments:
  {predict,lat_pred,get_ir,create,connect,register,unregister}
    predict (lat_pred)  apply latency predictor for testing model
    get_ir              specify a model type to convert to nn-meter ir graph
    create              create a workspace folder for nn-Meter builder
    connect             connect to backend
    register            register customized module to nn-Meter, supporting
                        type: predictor, backend, operator, testcase, operator
    unregister          unregister customized module from nn-Meter, supporting
                        type: predictor, backend, operator, testcase, operator

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         increase output verbosity
  --list-predictors     list all supported predictors
  --list-backends       list all supported backends
  --list-kernels        list all supported kernels when building kernel
                        predictors
  --list-operators      list all supported operators when building fusion rule
                        test cases
  --list-testcases      list all supported special test cases when building
                        fusion rule test cases
(py37) yongqiang@yongqiang:~$

2.3. Sample models

/home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/

(py37) yongqiang@yongqiang:~/yongqiang_work/nn-Meter/material/testmodels$ pwd
/home/yongqiang/yongqiang_work/nn-Meter/material/testmodels
(py37) yongqiang@yongqiang:~/yongqiang_work/nn-Meter/material/testmodels$ ls -l
total 18080
-rw-r--r-- 1 yongqiang yongqiang   239602 Nov 12 10:13 mobilenetv3small_0.json
-rw-r--r-- 1 yongqiang yongqiang 10169165 Nov 12 10:13 mobilenetv3small_0.onnx
-rw-r--r-- 1 yongqiang yongqiang  8098911 Nov 12 10:13 mobilenetv3small_0.pb
(py37) yongqiang@yongqiang:~/yongqiang_work/nn-Meter/material/testmodels$

2.4. v1.0-data and v2.0-data

https://github.com/microsoft/nn-Meter/releases/tag/v1.0-data

adreno630gpu_tflite21.zip
adreno640gpu_tflite21.zip
cortexA76cpu_tflite21.zip
datasets.zip
ir_graphs.zip
myriadvpu_openvino2019r2.zip
onnx_models.zip
pb_models.zip
Source code (zip)
Source code (tar.gz)

https://github.com/microsoft/nn-Meter/releases/tag/v2.0-data

tflite_benchmark_tools_v2.1.zip
tflite_benchmark_tools_v2.7.zip
Source code (zip)
Source code (tar.gz)

~/.nn_meter/config

(py37) yongqiang@yongqiang:~/.nn_meter/config$ pwd
/home/yongqiang/.nn_meter/config
(py37) yongqiang@yongqiang:~/.nn_meter/config$
(py37) yongqiang@yongqiang:~/.nn_meter/config$ ls -l
total 8
-rw-r--r-- 1 yongqiang yongqiang 1414 Nov 12 15:04 predictors.yaml
-rw-r--r-- 1 yongqiang yongqiang   44 Nov 12 15:04 settings.yaml
(py37) yongqiang@yongqiang:~/.nn_meter/config$ cat predictors.yaml
- name: cortexA76cpu_tflite21
  version: 1.0
  category: cpu
  kernel_predictors:
    - conv-bn-relu
    - dwconv-bn-relu
    - fc
    - global-avgpool
    - hswish
    - relu
    - se
    - split
    - add
    - addrelu
    - maxpool
    - avgpool
    - bn
    - bnrelu
    - channelshuffle
    - concat
  download: https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/cortexA76cpu_tflite21.zip

- name: adreno640gpu_tflite21
  version: 1.0
  category: gpu
  kernel_predictors:
    - conv-bn-relu
    - dwconv-bn-relu
    - hswish
    - relu
    - se
    - maxpool
    - avgpool
    - bn
    - bnrelu
    - concat
  download: https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/adreno640gpu_tflite21.zip

- name: adreno630gpu_tflite21
  version: 1.0
  category: gpu
  kernel_predictors:
    - conv-bn-relu
    - dwconv-bn-relu
    - hswish
    - relu
    - se
    - maxpool
    - avgpool
    - bn
    - bnrelu
    - concat
  download: https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/adreno630gpu_tflite21.zip

- name: myriadvpu_openvino2019r2
  version: 1.0
  category: vpu
  kernel_predictors:
    - conv-bn-relu
    - dwconv-bn-relu
    - fc
    - hswish
    - relu
    - se
    - addrelu
    - maxpool
    - avgpool
    - bn
    - bnrelu
    - channelshuffle
  download: https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/myriadvpu_openvino2019r2.zip
(py37) yongqiang@yongqiang:~/.nn_meter/config$

data_folder: /home/yongqiang/.nn_meter/data

(py37) yongqiang@yongqiang:~/.nn_meter/config$ cat settings.yaml
data_folder: /home/yongqiang/.nn_meter/data
(py37) yongqiang@yongqiang:~/.nn_meter/config$

nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) Download from https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/cortexA76cpu_tflite21.zip ...
  0%|▎                                                      | 868k/376M [14:29<66:56:25, 1.56kiB/s]
  Traceback (most recent call last):
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/urllib3/response.py", line 710, in _error_catcher
    yield
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/urllib3/response.py", line 835, in _raw_read
    raise IncompleteRead(self._fp_bytes_read, self.length_remaining)
urllib3.exceptions.IncompleteRead: IncompleteRead(884149 bytes read, 375517932 more expected)

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/requests/models.py", line 816, in generate
    yield from self.raw.stream(chunk_size, decode_content=True)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/urllib3/response.py", line 936, in stream
    data = self.read(amt=amt, decode_content=decode_content)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/urllib3/response.py", line 907, in read
    data = self._raw_read(amt)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/urllib3/response.py", line 835, in _raw_read
    raise IncompleteRead(self._fp_bytes_read, self.length_remaining)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/contextlib.py", line 130, in __exit__
    self.gen.throw(type, value, traceback)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/urllib3/response.py", line 727, in _error_catcher
    raise ProtocolError(f"Connection broken: {e!r}", e) from e
urllib3.exceptions.ProtocolError: ('Connection broken: IncompleteRead(884149 bytes read, 375517932 more expected)', IncompleteRead(884149 bytes read, 375517932 more expected))

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/yongqiang/miniconda3/envs/py37/bin/nn-meter", line 8, in <module>
    sys.exit(nn_meter_cli())
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/interface.py", line 266, in nn_meter_cli
    args.func(args)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/predictor.py", line 39, in apply_latency_predictor_cli
    predictor = load_latency_predictor(args.predictor, args.predictor_version)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/predictor/nn_meter_predictor.py", line 66, in load_latency_predictor
    kernel_predictors, fusionrule = loading_to_local(pred_info, os.path.join(user_data_folder, 'predictor'))
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/predictor/utils.py", line 30, in loading_to_local
    download_from_url(pred_info["download"], dir)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/utils.py", line 29, in download_from_url
    for data in response.iter_content(block_size):
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/requests/models.py", line 818, in generate
    raise ChunkedEncodingError(e)
requests.exceptions.ChunkedEncodingError: ('Connection broken: IncompleteRead(884149 bytes read, 375517932 more expected)', IncompleteRead(884149 bytes read, 375517932 more expected))
(py37) yongqiang@yongqiang:~$

在线下载 https://github.com/microsoft/nn-Meter/releases/download/v1.0-data/cortexA76cpu_tflite21.zip 失败，可以从 https://github.com/microsoft/nn-Meter/releases/tag/v1.0-data 处离线下载，复制到 data_folder: /home/yongqiang/.nn_meter/data 目录，然后解压，即可使用。

(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$ chmod a+x cortexA76cpu_tflite21.zip
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$ ls -l
total 367584
-rwxr-xr-x 1 yongqiang yongqiang 376402081 Nov 19 21:06 cortexA76cpu_tflite21.zip
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$ unzip cortexA76cpu_tflite21.zip
Archive:  cortexA76cpu_tflite21.zip
   creating: cortexA76cpu_tflite21/
  inflating: cortexA76cpu_tflite21/add.pkl
  inflating: cortexA76cpu_tflite21/addrelu.pkl
  inflating: cortexA76cpu_tflite21/avgpool.pkl
  inflating: cortexA76cpu_tflite21/bn.pkl
  inflating: cortexA76cpu_tflite21/bnrelu.pkl
  inflating: cortexA76cpu_tflite21/channelshuffle.pkl
  inflating: cortexA76cpu_tflite21/concat.pkl
  inflating: cortexA76cpu_tflite21/conv-bn-relu.pkl
  inflating: cortexA76cpu_tflite21/dwconv-bn-relu.pkl
  inflating: cortexA76cpu_tflite21/fc.pkl
  inflating: cortexA76cpu_tflite21/fusion_rules.json
  inflating: cortexA76cpu_tflite21/global-avgpool.pkl
  inflating: cortexA76cpu_tflite21/hswish.pkl
  inflating: cortexA76cpu_tflite21/maxpool.pkl
  inflating: cortexA76cpu_tflite21/relu.pkl
  inflating: cortexA76cpu_tflite21/se.pkl
  inflating: cortexA76cpu_tflite21/split.pkl
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$ ls -l
total 367588
drwxr-xr-x 2 yongqiang yongqiang      4096 Jul 23  2021 cortexA76cpu_tflite21
-rwxr-xr-x 1 yongqiang yongqiang 376402081 Nov 19 21:06 cortexA76cpu_tflite21.zip
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor$ cd cortexA76cpu_tflite21/
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor/cortexA76cpu_tflite21$ ls -l
total 1369956
-rw-r--r-- 1 yongqiang yongqiang  31757462 May 31  2021 add.pkl
-rw-r--r-- 1 yongqiang yongqiang  31757462 May 31  2021 addrelu.pkl
-rw-r--r-- 1 yongqiang yongqiang  77011080 May 31  2021 avgpool.pkl
-rw-r--r-- 1 yongqiang yongqiang  30081902 May 31  2021 bn.pkl
-rw-r--r-- 1 yongqiang yongqiang  31535606 May 31  2021 bnrelu.pkl
-rw-r--r-- 1 yongqiang yongqiang  14827163 May 31  2021 channelshuffle.pkl
-rw-r--r-- 1 yongqiang yongqiang 335113694 May 31  2021 concat.pkl
-rw-r--r-- 1 yongqiang yongqiang 525120135 May 31  2021 conv-bn-relu.pkl
-rw-r--r-- 1 yongqiang yongqiang  82660737 May 31  2021 dwconv-bn-relu.pkl
-rw-r--r-- 1 yongqiang yongqiang  58072608 May 31  2021 fc.pkl
-rw-r--r-- 1 yongqiang yongqiang     16421 Jun  2  2021 fusion_rules.json
-rw-r--r-- 1 yongqiang yongqiang  28156904 May 31  2021 global-avgpool.pkl
-rw-r--r-- 1 yongqiang yongqiang  24177043 May 31  2021 hswish.pkl
-rw-r--r-- 1 yongqiang yongqiang  10680143 May 31  2021 maxpool.pkl
-rw-r--r-- 1 yongqiang yongqiang  31535606 May 31  2021 relu.pkl
-rw-r--r-- 1 yongqiang yongqiang  60680510 May 31  2021 se.pkl
-rw-r--r-- 1 yongqiang yongqiang  29598568 May 31  2021 split.pkl
(py37) yongqiang@yongqiang:~/.nn_meter/data/predictor/cortexA76cpu_tflite21$

2.5. scikit-learn==0.23.1

conda install scikit-learn==0.23.1

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/sklearn/base.py:338: UserWarning: Trying to unpickle estimator DecisionTreeRegressor from version 0.23.1 when using version 1.0.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/modules/model_persistence.html#security-maintainability-limitations
  UserWarning,
/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/sklearn/base.py:338: UserWarning: Trying to unpickle estimator RandomForestRegressor from version 0.23.1 when using version 1.0.2. This might lead to breaking code or invalid results. Use at your own risk. For more info please refer to:
https://scikit-learn.org/stable/modules/model_persistence.html#security-maintainability-limitations
  UserWarning,
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
(nn-Meter) You have not install the tensorflow package, please install tensorflow==2.7.0 and try again.
Traceback (most recent call last):
  File "/home/yongqiang/miniconda3/envs/py37/bin/nn-meter", line 8, in <module>
    sys.exit(nn_meter_cli())
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/interface.py", line 266, in nn_meter_cli
    args.func(args)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/predictor.py", line 56, in apply_latency_predictor_cli
    latency = predictor.predict(model, model_type) # in unit of ms
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/predictor/nn_meter_predictor.py", line 106, in predict
    graph = model_file_to_graph(model, model_type, input_shape, apply_nni=apply_nni)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/utils.py", line 42, in model_file_to_graph
    converter = FrozenPbConverter(filename)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/frozenpb_converter/frozenpb_converter.py", line 15, in __init__
    parser = FrozenPbParser(file_name)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/frozenpb_converter/frozenpb_parser.py", line 15, in __init__
    graph = tf.compat.v1.GraphDef()
AttributeError: 'NoneType' object has no attribute 'compat'
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ pip uninstall scikit-learn
Found existing installation: scikit-learn 1.0.2
Uninstalling scikit-learn-1.0.2:
  Would remove:
    /home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/scikit_learn-1.0.2.dist-info/*
    /home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/scikit_learn.libs/libgomp-a34b3233.so.1.0.0
    /home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/sklearn/*
Proceed (Y/n)? y
  Successfully uninstalled scikit-learn-1.0.2
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ conda install scikit-learn==0.23.1
Collecting package metadata (current_repodata.json): done
...
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ conda list | grep scikit-learn
scikit-learn              0.23.1           py37h423224d_0    defaults
(py37) yongqiang@yongqiang:~$

2.6. tensorflow==2.7.0

pip install tensorflow==2.7.0 --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
(nn-Meter) You have not install the tensorflow package, please install tensorflow==2.7.0 and try again.
Traceback (most recent call last):
  File "/home/yongqiang/miniconda3/envs/py37/bin/nn-meter", line 8, in <module>
    sys.exit(nn_meter_cli())
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/interface.py", line 266, in nn_meter_cli
    args.func(args)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/predictor.py", line 56, in apply_latency_predictor_cli
    latency = predictor.predict(model, model_type) # in unit of ms
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/predictor/nn_meter_predictor.py", line 106, in predict
    graph = model_file_to_graph(model, model_type, input_shape, apply_nni=apply_nni)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/utils.py", line 42, in model_file_to_graph
    converter = FrozenPbConverter(filename)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/frozenpb_converter/frozenpb_converter.py", line 15, in __init__
    parser = FrozenPbParser(file_name)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/frozenpb_converter/frozenpb_parser.py", line 15, in __init__
    graph = tf.compat.v1.GraphDef()
AttributeError: 'NoneType' object has no attribute 'compat'
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ pip install tensorflow==2.7.0 --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Collecting tensorflow==2.7.0
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/e8/02/d981f022c0232d692db870501c4198cb9de2d6f00b50ee84e58fa755ca84/tensorflow-2.7.0-cp37-cp37m-manylinux2010_x86_64.whl (489.6 MB)
...
(py37) yongqiang@yongqiang:~$

2.7. onnx==1.10.0

pip install onnx==1.10.0 --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --onnx /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.onnx
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
(nn-Meter) You have not install the onnx package, please install onnx==1.10.0 and try again.
Traceback (most recent call last):
  File "/home/yongqiang/miniconda3/envs/py37/bin/nn-meter", line 8, in <module>
    sys.exit(nn_meter_cli())
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/interface.py", line 266, in nn_meter_cli
    args.func(args)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/utils/nn_meter_cli/predictor.py", line 56, in apply_latency_predictor_cli
    latency = predictor.predict(model, model_type) # in unit of ms
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/predictor/nn_meter_predictor.py", line 106, in predict
    graph = model_file_to_graph(model, model_type, input_shape, apply_nni=apply_nni)
  File "/home/yongqiang/miniconda3/envs/py37/lib/python3.7/site-packages/nn_meter/ir_converter/utils.py", line 38, in model_file_to_graph
    model = onnx.load(filename)
AttributeError: 'NoneType' object has no attribute 'load'
(py37) yongqiang@yongqiang:~$
(py37) yongqiang@yongqiang:~$ pip install onnx==1.10.0 --default-timeout=1000 -i https://pypi.tuna.tsinghua.edu.cn/simple
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Collecting onnx==1.10.0
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/43/79/b274e2a918700012e0dcc0ae68f597110c3e45efbb5a7ce3b8569b3dd299/onnx-1.10.0-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (12.3 MB)
...
Installing collected packages: onnx
Successfully installed onnx-1.10.0
(py37) yongqiang@yongqiang:~$

3. Usage

3.1. Predict latency of saved CNN model

After installation, a command named nn-meter is enabled. To predict the latency for a CNN model with a predefined predictor in command line, users can run the following commands. Sample models can be downloaded here (https://github.com/microsoft/nn-Meter/tree/main/material/testmodels)

# for Tensorflow (*.pb) file
nn-meter predict --predictor <hardware> [--predictor-version <version>] --tensorflow <pb-file_or_folder> 
# Example Usage
nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb 

# for ONNX (*.onnx) file
nn-meter predict --predictor <hardware> [--predictor-version <version>] --onnx <onnx-file_or_folder>
#Example Usage
nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --onnx /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.onnx 

# for torch model from torchvision model zoo (str)
nn-meter predict --predictor <hardware> [--predictor-version <version>] --torchvision <model-name> <model-name>... 
#Example Usage
nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --torchvision /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenet_v2

# for nn-Meter IR (*.json) file
nn-meter predict --predictor <hardware> [--predictor-version <version>] --nn-meter-ir <json-file_or_folder> 
#Example Usage
nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --nn-meter-ir /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.json

--predictor-version <version> arguments is optional. When the predictor version is not specified by users, nn-meter will use the latest version of the predictor.

(base) yongqiang@yongqiang:~/yongqiang_work/nn-Meter/material/testmodels$ pwd
/home/yongqiang/yongqiang_work/nn-Meter/material/testmodels
(base) yongqiang@yongqiang:~/yongqiang_work/nn-Meter/material/testmodels$ ls -l
total 18080
-rw-r--r-- 1 yongqiang yongqiang   239602 Nov 12 10:13 mobilenetv3small_0.json
-rw-r--r-- 1 yongqiang yongqiang 10169165 Nov 12 10:13 mobilenetv3small_0.onnx
-rw-r--r-- 1 yongqiang yongqiang  8098911 Nov 12 10:13 mobilenetv3small_0.pb
(base) yongqiang@yongqiang:~/yongqiang_work/nn-Meter/material/testmodels$

(base) yongqiang@yongqiang:~/.nn_meter/config$ ll
total 16
drwxr-xr-x 2 yongqiang yongqiang 4096 Nov 12 15:04 ./
drwxr-xr-x 4 yongqiang yongqiang 4096 Nov 12 18:19 ../
-rw-r--r-- 1 yongqiang yongqiang 1414 Nov 12 15:04 predictors.yaml
-rw-r--r-- 1 yongqiang yongqiang   44 Nov 12 15:04 settings.yaml
(base) yongqiang@yongqiang:~/.nn_meter/config$
(base) yongqiang@yongqiang:~/.nn_meter/config$ cat settings.yaml
data_folder: /home/yongqiang/.nn_meter/data
(base) yongqiang@yongqiang:~/.nn_meter/config$

nn-Meter can support batch mode prediction. To predict latency for multiple models in the same model type once, user should collect all models in one folder and state the folder after --[model-type] liked argument.

It should also be noted that for PyTorch model, nn-meter can only support existing models in torchvision model zoo. The string followed by --torchvision should be exactly one or more string indicating name(s) of some existing torchvision models. To apply latency prediction for torchvision model in command line, onnx and onnx-simplifier packages are required.

nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
...
(nn-Meter) Predict latency: 12.558942703135 ms
(nn-Meter) [RESULT] predict latency for mobilenetv3small_0.pb: 12.558942703135 ms
(py37) yongqiang@yongqiang:~$

nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --onnx /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.onnx

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --onnx /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.onnx
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
(nn-Meter) Predict latency: 11.842191154847175 ms
(nn-Meter) [RESULT] predict latency for mobilenetv3small_0.onnx: 11.842191154847175 ms
(py37) yongqiang@yongqiang:~$

nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --nn-meter-ir /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.json

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --nn-meter-ir /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.json
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
(nn-Meter) Predict latency: 12.558942703135 ms
(nn-Meter) [RESULT] predict latency for mobilenetv3small_0.json: 12.558942703135 ms
(py37) yongqiang@yongqiang:~$

Appendix

nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb

(py37) yongqiang@yongqiang:~$ nn-meter predict --predictor cortexA76cpu_tflite21 --predictor-version 1.0 --tensorflow /home/yongqiang/yongqiang_work/nn-Meter/material/testmodels/mobilenetv3small_0.pb
(nn-Meter) checking local kernel predictors at /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/dwconv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/global-avgpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/concat.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/hswish.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/conv-bn-relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/channelshuffle.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/split.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/maxpool.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/fc.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/relu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/se.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bnrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/add.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/addrelu.pkl
(nn-Meter) load predictor /home/yongqiang/.nn_meter/data/predictor/cortexA76cpu_tflite21/bn.pkl
(nn-Meter) Start latency prediction ...
2023-11-20 23:22:53.764300: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /home/yongqiang/vulkan/1.3.239.0/x86_64/lib
2023-11-20 23:22:53.764628: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
(nn-Meter) Input shape of fc15.fc/weight op is [].
(nn-Meter) Output shape of fc15.fc/weight op is [[1024, 1000]].
(nn-Meter) Input shape of fc15.fc/weight/read op is [].
(nn-Meter) Output shape of fc15.fc/weight/read op is [[1024, 1000]].
(nn-Meter) Input shape of Reshape/shape op is [].
(nn-Meter) Output shape of Reshape/shape op is [[2]].
(nn-Meter) Input shape of Mean/reduction_indices op is [].
(nn-Meter) Output shape of Mean/reduction_indices op is [[2]].
(nn-Meter) Input shape of conv13.1.hswish/mul/y op is [].
(nn-Meter) Output shape of conv13.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of conv13.1.hswish/add/y op is [].
(nn-Meter) Output shape of conv13.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/moving_variance op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/moving_variance/read op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/moving_mean op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/moving_mean/read op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/gamma op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/gamma/read op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/beta op is [[1024]].
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/beta/read op is [[1024]].
(nn-Meter) Input shape of conv13.1.conv/weight op is [].
(nn-Meter) Output shape of conv13.1.conv/weight op is [[1, 1, 96, 1024]].
(nn-Meter) Input shape of conv13.1.conv/weight/read op is [].
(nn-Meter) Output shape of conv13.1.conv/weight/read op is [[1, 1, 96, 1024]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/moving_variance op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/moving_variance/read op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/moving_mean op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/moving_mean/read op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/gamma op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/gamma/read op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/beta op is [[96]].
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/beta/read op is [[96]].
(nn-Meter) Input shape of layer12.3.conv/weight op is [].
(nn-Meter) Output shape of layer12.3.conv/weight op is [[1, 1, 576, 96]].
(nn-Meter) Input shape of layer12.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer12.3.conv/weight/read op is [[1, 1, 576, 96]].
(nn-Meter) Input shape of layer12.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer12.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer12.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer12.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_16/y op is [].
(nn-Meter) Output shape of mul_16/y op is [[]].
(nn-Meter) Input shape of Add_13/y op is [].
(nn-Meter) Output shape of Add_13/y op is [[]].
(nn-Meter) Input shape of SE_8/conv2d_17/bias op is [].
(nn-Meter) Output shape of SE_8/conv2d_17/bias op is [[576]].
(nn-Meter) Input shape of SE_8/conv2d_17/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_8/conv2d_17/BiasAdd/ReadVariableOp op is [[576]].
(nn-Meter) Input shape of SE_8/conv2d_17/kernel op is [].
(nn-Meter) Output shape of SE_8/conv2d_17/kernel op is [[1, 1, 144, 576]].
(nn-Meter) Input shape of SE_8/conv2d_17/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_8/conv2d_17/Conv2D/ReadVariableOp op is [[1, 1, 144, 576]].
(nn-Meter) Input shape of SE_8/conv2d_16/bias op is [].
(nn-Meter) Output shape of SE_8/conv2d_16/bias op is [[144]].
(nn-Meter) Input shape of SE_8/conv2d_16/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_8/conv2d_16/BiasAdd/ReadVariableOp op is [[144]].
(nn-Meter) Input shape of SE_8/conv2d_16/kernel op is [].
(nn-Meter) Output shape of SE_8/conv2d_16/kernel op is [[1, 1, 576, 144]].
(nn-Meter) Input shape of SE_8/conv2d_16/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_8/conv2d_16/Conv2D/ReadVariableOp op is [[1, 1, 576, 144]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/moving_variance op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/moving_variance/read op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/moving_mean op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/moving_mean/read op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/gamma op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/gamma/read op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/beta op is [[576]].
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/beta/read op is [[576]].
(nn-Meter) Input shape of layer12.2.depconv/weight op is [].
(nn-Meter) Output shape of layer12.2.depconv/weight op is [[5, 5, 576, 1]].
(nn-Meter) Input shape of layer12.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer12.2.depconv/weight/read op is [[5, 5, 576, 1]].
(nn-Meter) Input shape of layer12.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer12.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer12.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer12.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/moving_variance op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/moving_variance/read op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/moving_mean op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/moving_mean/read op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/gamma op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/gamma/read op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/beta op is [[576]].
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/beta/read op is [[576]].
(nn-Meter) Input shape of layer12.1.conv/weight op is [].
(nn-Meter) Output shape of layer12.1.conv/weight op is [[1, 1, 96, 576]].
(nn-Meter) Input shape of layer12.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer12.1.conv/weight/read op is [[1, 1, 96, 576]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/moving_variance op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/moving_variance/read op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/moving_mean op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/moving_mean/read op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/gamma op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/gamma/read op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/beta op is [[96]].
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/beta/read op is [[96]].
(nn-Meter) Input shape of layer11.3.conv/weight op is [].
(nn-Meter) Output shape of layer11.3.conv/weight op is [[1, 1, 576, 96]].
(nn-Meter) Input shape of layer11.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer11.3.conv/weight/read op is [[1, 1, 576, 96]].
(nn-Meter) Input shape of layer11.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer11.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer11.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer11.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_14/y op is [].
(nn-Meter) Output shape of mul_14/y op is [[]].
(nn-Meter) Input shape of Add_11/y op is [].
(nn-Meter) Output shape of Add_11/y op is [[]].
(nn-Meter) Input shape of SE_7/conv2d_15/bias op is [].
(nn-Meter) Output shape of SE_7/conv2d_15/bias op is [[576]].
(nn-Meter) Input shape of SE_7/conv2d_15/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_7/conv2d_15/BiasAdd/ReadVariableOp op is [[576]].
(nn-Meter) Input shape of SE_7/conv2d_15/kernel op is [].
(nn-Meter) Output shape of SE_7/conv2d_15/kernel op is [[1, 1, 144, 576]].
(nn-Meter) Input shape of SE_7/conv2d_15/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_7/conv2d_15/Conv2D/ReadVariableOp op is [[1, 1, 144, 576]].
(nn-Meter) Input shape of SE_7/conv2d_14/bias op is [].
(nn-Meter) Output shape of SE_7/conv2d_14/bias op is [[144]].
(nn-Meter) Input shape of SE_7/conv2d_14/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_7/conv2d_14/BiasAdd/ReadVariableOp op is [[144]].
(nn-Meter) Input shape of SE_7/conv2d_14/kernel op is [].
(nn-Meter) Output shape of SE_7/conv2d_14/kernel op is [[1, 1, 576, 144]].
(nn-Meter) Input shape of SE_7/conv2d_14/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_7/conv2d_14/Conv2D/ReadVariableOp op is [[1, 1, 576, 144]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/moving_variance op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/moving_variance/read op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/moving_mean op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/moving_mean/read op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/gamma op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/gamma/read op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/beta op is [[576]].
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/beta/read op is [[576]].
(nn-Meter) Input shape of layer11.2.depconv/weight op is [].
(nn-Meter) Output shape of layer11.2.depconv/weight op is [[5, 5, 576, 1]].
(nn-Meter) Input shape of layer11.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer11.2.depconv/weight/read op is [[5, 5, 576, 1]].
(nn-Meter) Input shape of layer11.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer11.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer11.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer11.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/moving_variance op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/moving_variance/read op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/moving_mean op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/moving_mean/read op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/gamma op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/gamma/read op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/beta op is [[576]].
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/beta/read op is [[576]].
(nn-Meter) Input shape of layer11.1.conv/weight op is [].
(nn-Meter) Output shape of layer11.1.conv/weight op is [[1, 1, 96, 576]].
(nn-Meter) Input shape of layer11.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer11.1.conv/weight/read op is [[1, 1, 96, 576]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/moving_variance op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/moving_variance/read op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/moving_mean op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/moving_mean/read op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/gamma op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/gamma/read op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/beta op is [[96]].
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/beta/read op is [[96]].
(nn-Meter) Input shape of layer10.3.conv/weight op is [].
(nn-Meter) Output shape of layer10.3.conv/weight op is [[1, 1, 288, 96]].
(nn-Meter) Input shape of layer10.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer10.3.conv/weight/read op is [[1, 1, 288, 96]].
(nn-Meter) Input shape of layer10.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer10.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer10.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer10.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_12/y op is [].
(nn-Meter) Output shape of mul_12/y op is [[]].
(nn-Meter) Input shape of Add_10/y op is [].
(nn-Meter) Output shape of Add_10/y op is [[]].
(nn-Meter) Input shape of SE_6/conv2d_13/bias op is [].
(nn-Meter) Output shape of SE_6/conv2d_13/bias op is [[288]].
(nn-Meter) Input shape of SE_6/conv2d_13/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_6/conv2d_13/BiasAdd/ReadVariableOp op is [[288]].
(nn-Meter) Input shape of SE_6/conv2d_13/kernel op is [].
(nn-Meter) Output shape of SE_6/conv2d_13/kernel op is [[1, 1, 72, 288]].
(nn-Meter) Input shape of SE_6/conv2d_13/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_6/conv2d_13/Conv2D/ReadVariableOp op is [[1, 1, 72, 288]].
(nn-Meter) Input shape of SE_6/conv2d_12/bias op is [].
(nn-Meter) Output shape of SE_6/conv2d_12/bias op is [[72]].
(nn-Meter) Input shape of SE_6/conv2d_12/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_6/conv2d_12/BiasAdd/ReadVariableOp op is [[72]].
(nn-Meter) Input shape of SE_6/conv2d_12/kernel op is [].
(nn-Meter) Output shape of SE_6/conv2d_12/kernel op is [[1, 1, 288, 72]].
(nn-Meter) Input shape of SE_6/conv2d_12/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_6/conv2d_12/Conv2D/ReadVariableOp op is [[1, 1, 288, 72]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/moving_variance op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/moving_variance/read op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/moving_mean op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/moving_mean/read op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/gamma op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/gamma/read op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/beta op is [[288]].
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/beta/read op is [[288]].
(nn-Meter) Input shape of layer10.2.depconv/weight op is [].
(nn-Meter) Output shape of layer10.2.depconv/weight op is [[5, 5, 288, 1]].
(nn-Meter) Input shape of layer10.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer10.2.depconv/weight/read op is [[5, 5, 288, 1]].
(nn-Meter) Input shape of layer10.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer10.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer10.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer10.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/moving_variance op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/moving_variance/read op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/moving_mean op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/moving_mean/read op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/gamma op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/gamma/read op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/beta op is [[288]].
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/beta/read op is [[288]].
(nn-Meter) Input shape of layer10.1.conv/weight op is [].
(nn-Meter) Output shape of layer10.1.conv/weight op is [[1, 1, 48, 288]].
(nn-Meter) Input shape of layer10.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer10.1.conv/weight/read op is [[1, 1, 48, 288]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/moving_variance op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/moving_variance/read op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/moving_mean op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/moving_mean/read op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/gamma op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/gamma/read op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/beta op is [[48]].
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/beta/read op is [[48]].
(nn-Meter) Input shape of layer9.3.conv/weight op is [].
(nn-Meter) Output shape of layer9.3.conv/weight op is [[1, 1, 144, 48]].
(nn-Meter) Input shape of layer9.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer9.3.conv/weight/read op is [[1, 1, 144, 48]].
(nn-Meter) Input shape of layer9.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer9.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer9.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer9.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_10/y op is [].
(nn-Meter) Output shape of mul_10/y op is [[]].
(nn-Meter) Input shape of Add_8/y op is [].
(nn-Meter) Output shape of Add_8/y op is [[]].
(nn-Meter) Input shape of SE_5/conv2d_11/bias op is [].
(nn-Meter) Output shape of SE_5/conv2d_11/bias op is [[144]].
(nn-Meter) Input shape of SE_5/conv2d_11/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_5/conv2d_11/BiasAdd/ReadVariableOp op is [[144]].
(nn-Meter) Input shape of SE_5/conv2d_11/kernel op is [].
(nn-Meter) Output shape of SE_5/conv2d_11/kernel op is [[1, 1, 36, 144]].
(nn-Meter) Input shape of SE_5/conv2d_11/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_5/conv2d_11/Conv2D/ReadVariableOp op is [[1, 1, 36, 144]].
(nn-Meter) Input shape of SE_5/conv2d_10/bias op is [].
(nn-Meter) Output shape of SE_5/conv2d_10/bias op is [[36]].
(nn-Meter) Input shape of SE_5/conv2d_10/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_5/conv2d_10/BiasAdd/ReadVariableOp op is [[36]].
(nn-Meter) Input shape of SE_5/conv2d_10/kernel op is [].
(nn-Meter) Output shape of SE_5/conv2d_10/kernel op is [[1, 1, 144, 36]].
(nn-Meter) Input shape of SE_5/conv2d_10/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_5/conv2d_10/Conv2D/ReadVariableOp op is [[1, 1, 144, 36]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/moving_variance op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/moving_variance/read op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/moving_mean op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/moving_mean/read op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/gamma op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/gamma/read op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/beta op is [[144]].
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/beta/read op is [[144]].
(nn-Meter) Input shape of layer9.2.depconv/weight op is [].
(nn-Meter) Output shape of layer9.2.depconv/weight op is [[5, 5, 144, 1]].
(nn-Meter) Input shape of layer9.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer9.2.depconv/weight/read op is [[5, 5, 144, 1]].
(nn-Meter) Input shape of layer9.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer9.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer9.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer9.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/moving_variance op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/moving_variance/read op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/moving_mean op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/moving_mean/read op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/gamma op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/gamma/read op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/beta op is [[144]].
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/beta/read op is [[144]].
(nn-Meter) Input shape of layer9.1.conv/weight op is [].
(nn-Meter) Output shape of layer9.1.conv/weight op is [[1, 1, 48, 144]].
(nn-Meter) Input shape of layer9.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer9.1.conv/weight/read op is [[1, 1, 48, 144]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/moving_variance op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/moving_variance/read op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/moving_mean op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/moving_mean/read op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/gamma op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/gamma/read op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/beta op is [[48]].
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/beta/read op is [[48]].
(nn-Meter) Input shape of layer8.3.conv/weight op is [].
(nn-Meter) Output shape of layer8.3.conv/weight op is [[1, 1, 120, 48]].
(nn-Meter) Input shape of layer8.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer8.3.conv/weight/read op is [[1, 1, 120, 48]].
(nn-Meter) Input shape of layer8.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer8.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer8.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer8.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_8/y op is [].
(nn-Meter) Output shape of mul_8/y op is [[]].
(nn-Meter) Input shape of Add_7/y op is [].
(nn-Meter) Output shape of Add_7/y op is [[]].
(nn-Meter) Input shape of SE_4/conv2d_9/bias op is [].
(nn-Meter) Output shape of SE_4/conv2d_9/bias op is [[120]].
(nn-Meter) Input shape of SE_4/conv2d_9/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_4/conv2d_9/BiasAdd/ReadVariableOp op is [[120]].
(nn-Meter) Input shape of SE_4/conv2d_9/kernel op is [].
(nn-Meter) Output shape of SE_4/conv2d_9/kernel op is [[1, 1, 30, 120]].
(nn-Meter) Input shape of SE_4/conv2d_9/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_4/conv2d_9/Conv2D/ReadVariableOp op is [[1, 1, 30, 120]].
(nn-Meter) Input shape of SE_4/conv2d_8/bias op is [].
(nn-Meter) Output shape of SE_4/conv2d_8/bias op is [[30]].
(nn-Meter) Input shape of SE_4/conv2d_8/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_4/conv2d_8/BiasAdd/ReadVariableOp op is [[30]].
(nn-Meter) Input shape of SE_4/conv2d_8/kernel op is [].
(nn-Meter) Output shape of SE_4/conv2d_8/kernel op is [[1, 1, 120, 30]].
(nn-Meter) Input shape of SE_4/conv2d_8/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_4/conv2d_8/Conv2D/ReadVariableOp op is [[1, 1, 120, 30]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/moving_variance op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/moving_variance/read op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/moving_mean op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/moving_mean/read op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/gamma op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/gamma/read op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/beta op is [[120]].
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/beta/read op is [[120]].
(nn-Meter) Input shape of layer8.2.depconv/weight op is [].
(nn-Meter) Output shape of layer8.2.depconv/weight op is [[5, 5, 120, 1]].
(nn-Meter) Input shape of layer8.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer8.2.depconv/weight/read op is [[5, 5, 120, 1]].
(nn-Meter) Input shape of layer8.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer8.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer8.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer8.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/moving_variance op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/moving_variance/read op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/moving_mean op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/moving_mean/read op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/gamma op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/gamma/read op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/beta op is [[120]].
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/beta/read op is [[120]].
(nn-Meter) Input shape of layer8.1.conv/weight op is [].
(nn-Meter) Output shape of layer8.1.conv/weight op is [[1, 1, 40, 120]].
(nn-Meter) Input shape of layer8.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer8.1.conv/weight/read op is [[1, 1, 40, 120]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/moving_variance op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/moving_variance/read op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/moving_mean op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/moving_mean/read op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/gamma op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/gamma/read op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/beta op is [[40]].
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/beta/read op is [[40]].
(nn-Meter) Input shape of layer7.3.conv/weight op is [].
(nn-Meter) Output shape of layer7.3.conv/weight op is [[1, 1, 240, 40]].
(nn-Meter) Input shape of layer7.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer7.3.conv/weight/read op is [[1, 1, 240, 40]].
(nn-Meter) Input shape of layer7.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer7.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer7.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer7.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_6/y op is [].
(nn-Meter) Output shape of mul_6/y op is [[]].
(nn-Meter) Input shape of Add_5/y op is [].
(nn-Meter) Output shape of Add_5/y op is [[]].
(nn-Meter) Input shape of SE_3/conv2d_7/bias op is [].
(nn-Meter) Output shape of SE_3/conv2d_7/bias op is [[240]].
(nn-Meter) Input shape of SE_3/conv2d_7/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_3/conv2d_7/BiasAdd/ReadVariableOp op is [[240]].
(nn-Meter) Input shape of SE_3/conv2d_7/kernel op is [].
(nn-Meter) Output shape of SE_3/conv2d_7/kernel op is [[1, 1, 60, 240]].
(nn-Meter) Input shape of SE_3/conv2d_7/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_3/conv2d_7/Conv2D/ReadVariableOp op is [[1, 1, 60, 240]].
(nn-Meter) Input shape of SE_3/conv2d_6/bias op is [].
(nn-Meter) Output shape of SE_3/conv2d_6/bias op is [[60]].
(nn-Meter) Input shape of SE_3/conv2d_6/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_3/conv2d_6/BiasAdd/ReadVariableOp op is [[60]].
(nn-Meter) Input shape of SE_3/conv2d_6/kernel op is [].
(nn-Meter) Output shape of SE_3/conv2d_6/kernel op is [[1, 1, 240, 60]].
(nn-Meter) Input shape of SE_3/conv2d_6/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_3/conv2d_6/Conv2D/ReadVariableOp op is [[1, 1, 240, 60]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/moving_variance op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/moving_variance/read op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/moving_mean op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/moving_mean/read op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/gamma op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/gamma/read op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/beta op is [[240]].
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/beta/read op is [[240]].
(nn-Meter) Input shape of layer7.2.depconv/weight op is [].
(nn-Meter) Output shape of layer7.2.depconv/weight op is [[5, 5, 240, 1]].
(nn-Meter) Input shape of layer7.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer7.2.depconv/weight/read op is [[5, 5, 240, 1]].
(nn-Meter) Input shape of layer7.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer7.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer7.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer7.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/moving_variance op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/moving_variance/read op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/moving_mean op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/moving_mean/read op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/gamma op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/gamma/read op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/beta op is [[240]].
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/beta/read op is [[240]].
(nn-Meter) Input shape of layer7.1.conv/weight op is [].
(nn-Meter) Output shape of layer7.1.conv/weight op is [[1, 1, 40, 240]].
(nn-Meter) Input shape of layer7.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer7.1.conv/weight/read op is [[1, 1, 40, 240]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/moving_variance op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/moving_variance/read op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/moving_mean op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/moving_mean/read op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/gamma op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/gamma/read op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/beta op is [[40]].
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/beta/read op is [[40]].
(nn-Meter) Input shape of layer6.3.conv/weight op is [].
(nn-Meter) Output shape of layer6.3.conv/weight op is [[1, 1, 240, 40]].
(nn-Meter) Input shape of layer6.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer6.3.conv/weight/read op is [[1, 1, 240, 40]].
(nn-Meter) Input shape of layer6.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer6.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer6.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer6.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_4/y op is [].
(nn-Meter) Output shape of mul_4/y op is [[]].
(nn-Meter) Input shape of Add_3/y op is [].
(nn-Meter) Output shape of Add_3/y op is [[]].
(nn-Meter) Input shape of SE_2/conv2d_5/bias op is [].
(nn-Meter) Output shape of SE_2/conv2d_5/bias op is [[240]].
(nn-Meter) Input shape of SE_2/conv2d_5/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_2/conv2d_5/BiasAdd/ReadVariableOp op is [[240]].
(nn-Meter) Input shape of SE_2/conv2d_5/kernel op is [].
(nn-Meter) Output shape of SE_2/conv2d_5/kernel op is [[1, 1, 60, 240]].
(nn-Meter) Input shape of SE_2/conv2d_5/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_2/conv2d_5/Conv2D/ReadVariableOp op is [[1, 1, 60, 240]].
(nn-Meter) Input shape of SE_2/conv2d_4/bias op is [].
(nn-Meter) Output shape of SE_2/conv2d_4/bias op is [[60]].
(nn-Meter) Input shape of SE_2/conv2d_4/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_2/conv2d_4/BiasAdd/ReadVariableOp op is [[60]].
(nn-Meter) Input shape of SE_2/conv2d_4/kernel op is [].
(nn-Meter) Output shape of SE_2/conv2d_4/kernel op is [[1, 1, 240, 60]].
(nn-Meter) Input shape of SE_2/conv2d_4/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_2/conv2d_4/Conv2D/ReadVariableOp op is [[1, 1, 240, 60]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/moving_variance op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/moving_variance/read op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/moving_mean op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/moving_mean/read op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/gamma op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/gamma/read op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/beta op is [[240]].
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/beta/read op is [[240]].
(nn-Meter) Input shape of layer6.2.depconv/weight op is [].
(nn-Meter) Output shape of layer6.2.depconv/weight op is [[5, 5, 240, 1]].
(nn-Meter) Input shape of layer6.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer6.2.depconv/weight/read op is [[5, 5, 240, 1]].
(nn-Meter) Input shape of layer6.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer6.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer6.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer6.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/moving_variance op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/moving_variance/read op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/moving_mean op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/moving_mean/read op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/gamma op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/gamma/read op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/beta op is [[240]].
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/beta/read op is [[240]].
(nn-Meter) Input shape of layer6.1.conv/weight op is [].
(nn-Meter) Output shape of layer6.1.conv/weight op is [[1, 1, 40, 240]].
(nn-Meter) Input shape of layer6.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer6.1.conv/weight/read op is [[1, 1, 40, 240]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/moving_variance op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/moving_variance/read op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/moving_mean op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/moving_mean/read op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/gamma op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/gamma/read op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/beta op is [[40]].
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/beta/read op is [[40]].
(nn-Meter) Input shape of layer5.3.conv/weight op is [].
(nn-Meter) Output shape of layer5.3.conv/weight op is [[1, 1, 96, 40]].
(nn-Meter) Input shape of layer5.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer5.3.conv/weight/read op is [[1, 1, 96, 40]].
(nn-Meter) Input shape of layer5.2.hswish/mul/y op is [].
(nn-Meter) Output shape of layer5.2.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer5.2.hswish/add/y op is [].
(nn-Meter) Output shape of layer5.2.hswish/add/y op is [[]].
(nn-Meter) Input shape of mul_2/y op is [].
(nn-Meter) Output shape of mul_2/y op is [[]].
(nn-Meter) Input shape of Add_2/y op is [].
(nn-Meter) Output shape of Add_2/y op is [[]].
(nn-Meter) Input shape of SE_1/conv2d_3/bias op is [].
(nn-Meter) Output shape of SE_1/conv2d_3/bias op is [[96]].
(nn-Meter) Input shape of SE_1/conv2d_3/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_1/conv2d_3/BiasAdd/ReadVariableOp op is [[96]].
(nn-Meter) Input shape of SE_1/conv2d_3/kernel op is [].
(nn-Meter) Output shape of SE_1/conv2d_3/kernel op is [[1, 1, 24, 96]].
(nn-Meter) Input shape of SE_1/conv2d_3/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_1/conv2d_3/Conv2D/ReadVariableOp op is [[1, 1, 24, 96]].
(nn-Meter) Input shape of SE_1/conv2d_2/bias op is [].
(nn-Meter) Output shape of SE_1/conv2d_2/bias op is [[24]].
(nn-Meter) Input shape of SE_1/conv2d_2/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_1/conv2d_2/BiasAdd/ReadVariableOp op is [[24]].
(nn-Meter) Input shape of SE_1/conv2d_2/kernel op is [].
(nn-Meter) Output shape of SE_1/conv2d_2/kernel op is [[1, 1, 96, 24]].
(nn-Meter) Input shape of SE_1/conv2d_2/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE_1/conv2d_2/Conv2D/ReadVariableOp op is [[1, 1, 96, 24]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/moving_variance op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/moving_variance/read op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/moving_mean op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/moving_mean/read op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/gamma op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/gamma/read op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/beta op is [[96]].
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/beta/read op is [[96]].
(nn-Meter) Input shape of layer5.2.depconv/weight op is [].
(nn-Meter) Output shape of layer5.2.depconv/weight op is [[5, 5, 96, 1]].
(nn-Meter) Input shape of layer5.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer5.2.depconv/weight/read op is [[5, 5, 96, 1]].
(nn-Meter) Input shape of layer5.1.hswish/mul/y op is [].
(nn-Meter) Output shape of layer5.1.hswish/mul/y op is [[]].
(nn-Meter) Input shape of layer5.1.hswish/add/y op is [].
(nn-Meter) Output shape of layer5.1.hswish/add/y op is [[]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/moving_variance op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/moving_variance/read op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/moving_mean op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/moving_mean/read op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/gamma op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/gamma/read op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/beta op is [[96]].
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/beta/read op is [[96]].
(nn-Meter) Input shape of layer5.1.conv/weight op is [].
(nn-Meter) Output shape of layer5.1.conv/weight op is [[1, 1, 24, 96]].
(nn-Meter) Input shape of layer5.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer5.1.conv/weight/read op is [[1, 1, 24, 96]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/moving_variance op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/moving_variance/read op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/moving_mean op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/moving_mean/read op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/gamma op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/gamma/read op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/beta op is [[24]].
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/beta/read op is [[24]].
(nn-Meter) Input shape of layer4.3.conv/weight op is [].
(nn-Meter) Output shape of layer4.3.conv/weight op is [[1, 1, 88, 24]].
(nn-Meter) Input shape of layer4.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer4.3.conv/weight/read op is [[1, 1, 88, 24]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/moving_variance op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/moving_variance/read op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/moving_mean op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/moving_mean/read op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/gamma op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/gamma/read op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/beta op is [[88]].
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/beta/read op is [[88]].
(nn-Meter) Input shape of layer4.2.depconv/weight op is [].
(nn-Meter) Output shape of layer4.2.depconv/weight op is [[3, 3, 88, 1]].
(nn-Meter) Input shape of layer4.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer4.2.depconv/weight/read op is [[3, 3, 88, 1]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/moving_variance op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/moving_variance/read op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/moving_mean op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/moving_mean/read op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/gamma op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/gamma/read op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/beta op is [[88]].
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/beta/read op is [[88]].
(nn-Meter) Input shape of layer4.1.conv/weight op is [].
(nn-Meter) Output shape of layer4.1.conv/weight op is [[1, 1, 24, 88]].
(nn-Meter) Input shape of layer4.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer4.1.conv/weight/read op is [[1, 1, 24, 88]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/moving_variance op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/moving_variance/read op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/moving_mean op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/moving_mean/read op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/gamma op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/gamma/read op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/beta op is [[24]].
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/beta/read op is [[24]].
(nn-Meter) Input shape of layer3.3.conv/weight op is [].
(nn-Meter) Output shape of layer3.3.conv/weight op is [[1, 1, 72, 24]].
(nn-Meter) Input shape of layer3.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer3.3.conv/weight/read op is [[1, 1, 72, 24]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/moving_variance op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/moving_variance/read op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/moving_mean op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/moving_mean/read op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/gamma op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/gamma/read op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/beta op is [[72]].
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/beta/read op is [[72]].
(nn-Meter) Input shape of layer3.2.depconv/weight op is [].
(nn-Meter) Output shape of layer3.2.depconv/weight op is [[3, 3, 72, 1]].
(nn-Meter) Input shape of layer3.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer3.2.depconv/weight/read op is [[3, 3, 72, 1]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/moving_variance op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/moving_variance/read op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/moving_mean op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/moving_mean/read op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/gamma op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/gamma/read op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/beta op is [[72]].
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/beta/read op is [[72]].
(nn-Meter) Input shape of layer3.1.conv/weight op is [].
(nn-Meter) Output shape of layer3.1.conv/weight op is [[1, 1, 16, 72]].
(nn-Meter) Input shape of layer3.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer3.1.conv/weight/read op is [[1, 1, 16, 72]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/moving_variance op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/moving_variance/read op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/moving_mean op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/moving_mean/read op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/gamma op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/gamma/read op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/beta op is [[16]].
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/beta/read op is [[16]].
(nn-Meter) Input shape of layer2.3.conv/weight op is [].
(nn-Meter) Output shape of layer2.3.conv/weight op is [[1, 1, 16, 16]].
(nn-Meter) Input shape of layer2.3.conv/weight/read op is [].
(nn-Meter) Output shape of layer2.3.conv/weight/read op is [[1, 1, 16, 16]].
(nn-Meter) Input shape of mul/y op is [].
(nn-Meter) Output shape of mul/y op is [[]].
(nn-Meter) Input shape of Add/y op is [].
(nn-Meter) Output shape of Add/y op is [[]].
(nn-Meter) Input shape of SE/conv2d_1/bias op is [].
(nn-Meter) Output shape of SE/conv2d_1/bias op is [[16]].
(nn-Meter) Input shape of SE/conv2d_1/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE/conv2d_1/BiasAdd/ReadVariableOp op is [[16]].
(nn-Meter) Input shape of SE/conv2d_1/kernel op is [].
(nn-Meter) Output shape of SE/conv2d_1/kernel op is [[1, 1, 4, 16]].
(nn-Meter) Input shape of SE/conv2d_1/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE/conv2d_1/Conv2D/ReadVariableOp op is [[1, 1, 4, 16]].
(nn-Meter) Input shape of SE/conv2d/bias op is [].
(nn-Meter) Output shape of SE/conv2d/bias op is [[4]].
(nn-Meter) Input shape of SE/conv2d/BiasAdd/ReadVariableOp op is [].
(nn-Meter) Output shape of SE/conv2d/BiasAdd/ReadVariableOp op is [[4]].
(nn-Meter) Input shape of SE/conv2d/kernel op is [].
(nn-Meter) Output shape of SE/conv2d/kernel op is [[1, 1, 16, 4]].
(nn-Meter) Input shape of SE/conv2d/Conv2D/ReadVariableOp op is [].
(nn-Meter) Output shape of SE/conv2d/Conv2D/ReadVariableOp op is [[1, 1, 16, 4]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/moving_variance op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/moving_variance/read op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/moving_mean op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/moving_mean/read op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/gamma op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/gamma/read op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/beta op is [[16]].
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/beta/read op is [[16]].
(nn-Meter) Input shape of layer2.2.depconv/weight op is [].
(nn-Meter) Output shape of layer2.2.depconv/weight op is [[3, 3, 16, 1]].
(nn-Meter) Input shape of layer2.2.depconv/weight/read op is [].
(nn-Meter) Output shape of layer2.2.depconv/weight/read op is [[3, 3, 16, 1]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/moving_variance op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/moving_variance/read op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/moving_mean op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/moving_mean/read op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/gamma op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/gamma/read op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/beta op is [[16]].
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/beta/read op is [[16]].
(nn-Meter) Input shape of layer2.1.conv/weight op is [].
(nn-Meter) Output shape of layer2.1.conv/weight op is [[1, 1, 16, 16]].
(nn-Meter) Input shape of layer2.1.conv/weight/read op is [].
(nn-Meter) Output shape of layer2.1.conv/weight/read op is [[1, 1, 16, 16]].
(nn-Meter) Input shape of conv1.hswish.hswish/mul/y op is [].
(nn-Meter) Output shape of conv1.hswish.hswish/mul/y op is [[]].
(nn-Meter) Input shape of conv1.hswish.hswish/add/y op is [].
(nn-Meter) Output shape of conv1.hswish.hswish/add/y op is [[]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/moving_variance op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/moving_variance op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/moving_variance/read op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/moving_variance/read op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/moving_mean op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/moving_mean op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/moving_mean/read op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/moving_mean/read op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/gamma op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/gamma op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/gamma/read op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/gamma/read op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/beta op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/beta op is [[16]].
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/beta/read op is [].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/beta/read op is [[16]].
(nn-Meter) Input shape of conv1.conv/weight op is [].
(nn-Meter) Output shape of conv1.conv/weight op is [[3, 3, 3, 16]].
(nn-Meter) Input shape of conv1.conv/weight/read op is [].
(nn-Meter) Output shape of conv1.conv/weight/read op is [[3, 3, 3, 16]].
(nn-Meter) Input shape of input_im_0 op is [].
(nn-Meter) Output shape of input_im_0 op is [[1, 224, 224, 3]].
(nn-Meter) Find node conv1.conv/Conv2D with its weight op conv1.conv/weight.
(nn-Meter) Get input shape of conv1.conv/Conv2D from input_im_0, input shape:[1, 224, 224, 3].
(nn-Meter) Get weight shape of conv1.conv/Conv2D from ['conv1.conv/weight'], input shape:[3, 3, 3, 16].
(nn-Meter) Op:conv1.conv/Conv2D, stride:[1, 2, 2, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 224, 224, 3], kernel size: [2, 2], strides: [1, 2, 2, 1], padding: SAME.
(nn-Meter) Input shape of conv1.conv/Conv2D op is [[1, 224, 224, 3]].
(nn-Meter) Output shape of conv1.conv/Conv2D op is [[1, 112, 112, 16]].
(nn-Meter) Propagate through op conv1.bn.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of conv1.bn.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 112, 112, 16]].
(nn-Meter) Output shape of conv1.bn.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 112, 112, 16]].
(nn-Meter) Input shape of conv1.hswish.hswish/add op is [[1, 112, 112, 16], []].
(nn-Meter) Output shape of conv1.hswish.hswish/add op is [[1, 112, 112, 16]].
(nn-Meter) Propagate through op conv1.hswish.hswish/Relu6.
(nn-Meter) Input shape of conv1.hswish.hswish/Relu6 op is [[1, 112, 112, 16]].
(nn-Meter) Output shape of conv1.hswish.hswish/Relu6 op is [[1, 112, 112, 16]].
(nn-Meter) Input shape of conv1.hswish.hswish/mul op is [[1, 112, 112, 16], []].
(nn-Meter) Output shape of conv1.hswish.hswish/mul op is [[1, 112, 112, 16]].
(nn-Meter) Input shape of conv1.hswish.hswish/mul_1 op is [[1, 112, 112, 16], [1, 112, 112, 16]].
(nn-Meter) Output shape of conv1.hswish.hswish/mul_1 op is [[1, 112, 112, 16]].
(nn-Meter) Find node layer2.1.conv/Conv2D with its weight op layer2.1.conv/weight.
(nn-Meter) Get input shape of layer2.1.conv/Conv2D from conv1.hswish.hswish/mul_1, input shape:[1, 112, 112, 16].
(nn-Meter) Get weight shape of layer2.1.conv/Conv2D from ['layer2.1.conv/weight'], input shape:[1, 1, 16, 16].
(nn-Meter) Op:layer2.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 112, 112, 16], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer2.1.conv/Conv2D op is [[1, 112, 112, 16]].
(nn-Meter) Output shape of layer2.1.conv/Conv2D op is [[1, 112, 112, 16]].
(nn-Meter) Propagate through op layer2.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer2.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 112, 112, 16]].
(nn-Meter) Output shape of layer2.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 112, 112, 16]].
(nn-Meter) Propagate through op layer2.1.relu/Relu.
(nn-Meter) Input shape of layer2.1.relu/Relu op is [[1, 112, 112, 16]].
(nn-Meter) Output shape of layer2.1.relu/Relu op is [[1, 112, 112, 16]].
(nn-Meter) Find node layer2.2.depconv/depthwise with its weight op layer2.2.depconv/weight.
(nn-Meter) Get input shape of layer2.2.depconv/depthwise from layer2.1.relu/Relu, input shape:[1, 112, 112, 16].
(nn-Meter) Get weight shape of layer2.2.depconv/depthwise from ['layer2.2.depconv/weight'], input shape:[3, 3, 16, 1].
(nn-Meter) Op:layer2.2.depconv/depthwise, stride:[1, 2, 2, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 112, 112, 16], kernel size: [2, 2], strides: [1, 2, 2, 1], padding: SAME.
(nn-Meter) Input shape of layer2.2.depconv/depthwise op is [[1, 112, 112, 16]].
(nn-Meter) Output shape of layer2.2.depconv/depthwise op is [[1, 56, 56, 16]].
(nn-Meter) Propagate through op layer2.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer2.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 56, 56, 16]].
(nn-Meter) Output shape of layer2.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 56, 56, 16]].
(nn-Meter) Get input shape of SE/AvgPool from layer2.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 56, 56, 16].
(nn-Meter) Op:SE/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 56, 56, 16], kernel size: [1, 56, 56, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE/AvgPool op is [[1, 56, 56, 16]].
(nn-Meter) Output shape of SE/AvgPool op is [[1, 1, 1, 16]].
(nn-Meter) Find node SE/conv2d/Conv2D with its weight op SE/conv2d/kernel.
(nn-Meter) Get input shape of SE/conv2d/Conv2D from SE/AvgPool, input shape:[1, 1, 1, 16].
(nn-Meter) Get weight shape of SE/conv2d/Conv2D from ['SE/conv2d/kernel'], input shape:[1, 1, 16, 4].
(nn-Meter) Op:SE/conv2d/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 16], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE/conv2d/Conv2D op is [[1, 1, 1, 16]].
(nn-Meter) Output shape of SE/conv2d/Conv2D op is [[1, 1, 1, 4]].
(nn-Meter) Propagate through op SE/conv2d/BiasAdd.
(nn-Meter) Input shape of SE/conv2d/BiasAdd op is [[1, 1, 1, 4]].
(nn-Meter) Output shape of SE/conv2d/BiasAdd op is [[1, 1, 1, 4]].
(nn-Meter) Propagate through op SE/Relu.
(nn-Meter) Input shape of SE/Relu op is [[1, 1, 1, 4]].
(nn-Meter) Output shape of SE/Relu op is [[1, 1, 1, 4]].
(nn-Meter) Find node SE/conv2d_1/Conv2D with its weight op SE/conv2d_1/kernel.
(nn-Meter) Get input shape of SE/conv2d_1/Conv2D from SE/Relu, input shape:[1, 1, 1, 4].
(nn-Meter) Get weight shape of SE/conv2d_1/Conv2D from ['SE/conv2d_1/kernel'], input shape:[1, 1, 4, 16].
(nn-Meter) Op:SE/conv2d_1/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 4], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE/conv2d_1/Conv2D op is [[1, 1, 1, 4]].
(nn-Meter) Output shape of SE/conv2d_1/Conv2D op is [[1, 1, 1, 16]].
(nn-Meter) Propagate through op SE/conv2d_1/BiasAdd.
(nn-Meter) Input shape of SE/conv2d_1/BiasAdd op is [[1, 1, 1, 16]].
(nn-Meter) Output shape of SE/conv2d_1/BiasAdd op is [[1, 1, 1, 16]].
(nn-Meter) Input shape of Add op is [[1, 1, 1, 16], []].
(nn-Meter) Output shape of Add op is [[1, 1, 1, 16]].
(nn-Meter) Propagate through op Relu6.
(nn-Meter) Input shape of Relu6 op is [[1, 1, 1, 16]].
(nn-Meter) Output shape of Relu6 op is [[1, 1, 1, 16]].
(nn-Meter) Input shape of mul op is [[1, 1, 1, 16], []].
(nn-Meter) Output shape of mul op is [[1, 1, 1, 16]].
(nn-Meter) Input shape of mul_1 op is [[1, 56, 56, 16], [1, 56, 56, 16]].
(nn-Meter) Output shape of mul_1 op is [[1, 56, 56, 16]].
(nn-Meter) Propagate through op layer2.2.relu/Relu.
(nn-Meter) Input shape of layer2.2.relu/Relu op is [[1, 56, 56, 16]].
(nn-Meter) Output shape of layer2.2.relu/Relu op is [[1, 56, 56, 16]].
(nn-Meter) Find node layer2.3.conv/Conv2D with its weight op layer2.3.conv/weight.
(nn-Meter) Get input shape of layer2.3.conv/Conv2D from layer2.2.relu/Relu, input shape:[1, 56, 56, 16].
(nn-Meter) Get weight shape of layer2.3.conv/Conv2D from ['layer2.3.conv/weight'], input shape:[1, 1, 16, 16].
(nn-Meter) Op:layer2.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 56, 56, 16], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer2.3.conv/Conv2D op is [[1, 56, 56, 16]].
(nn-Meter) Output shape of layer2.3.conv/Conv2D op is [[1, 56, 56, 16]].
(nn-Meter) Propagate through op layer2.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer2.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 56, 56, 16]].
(nn-Meter) Output shape of layer2.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 56, 56, 16]].
(nn-Meter) Find node layer3.1.conv/Conv2D with its weight op layer3.1.conv/weight.
(nn-Meter) Get input shape of layer3.1.conv/Conv2D from layer2.3.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 56, 56, 16].
(nn-Meter) Get weight shape of layer3.1.conv/Conv2D from ['layer3.1.conv/weight'], input shape:[1, 1, 16, 72].
(nn-Meter) Op:layer3.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 56, 56, 16], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer3.1.conv/Conv2D op is [[1, 56, 56, 16]].
(nn-Meter) Output shape of layer3.1.conv/Conv2D op is [[1, 56, 56, 72]].
(nn-Meter) Propagate through op layer3.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer3.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 56, 56, 72]].
(nn-Meter) Output shape of layer3.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 56, 56, 72]].
(nn-Meter) Propagate through op layer3.1.relu/Relu.
(nn-Meter) Input shape of layer3.1.relu/Relu op is [[1, 56, 56, 72]].
(nn-Meter) Output shape of layer3.1.relu/Relu op is [[1, 56, 56, 72]].
(nn-Meter) Find node layer3.2.depconv/depthwise with its weight op layer3.2.depconv/weight.
(nn-Meter) Get input shape of layer3.2.depconv/depthwise from layer3.1.relu/Relu, input shape:[1, 56, 56, 72].
(nn-Meter) Get weight shape of layer3.2.depconv/depthwise from ['layer3.2.depconv/weight'], input shape:[3, 3, 72, 1].
(nn-Meter) Op:layer3.2.depconv/depthwise, stride:[1, 2, 2, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 56, 56, 72], kernel size: [2, 2], strides: [1, 2, 2, 1], padding: SAME.
(nn-Meter) Input shape of layer3.2.depconv/depthwise op is [[1, 56, 56, 72]].
(nn-Meter) Output shape of layer3.2.depconv/depthwise op is [[1, 28, 28, 72]].
(nn-Meter) Propagate through op layer3.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer3.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 72]].
(nn-Meter) Output shape of layer3.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 72]].
(nn-Meter) Propagate through op layer3.2.relu/Relu.
(nn-Meter) Input shape of layer3.2.relu/Relu op is [[1, 28, 28, 72]].
(nn-Meter) Output shape of layer3.2.relu/Relu op is [[1, 28, 28, 72]].
(nn-Meter) Find node layer3.3.conv/Conv2D with its weight op layer3.3.conv/weight.
(nn-Meter) Get input shape of layer3.3.conv/Conv2D from layer3.2.relu/Relu, input shape:[1, 28, 28, 72].
(nn-Meter) Get weight shape of layer3.3.conv/Conv2D from ['layer3.3.conv/weight'], input shape:[1, 1, 72, 24].
(nn-Meter) Op:layer3.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 28, 28, 72], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer3.3.conv/Conv2D op is [[1, 28, 28, 72]].
(nn-Meter) Output shape of layer3.3.conv/Conv2D op is [[1, 28, 28, 24]].
(nn-Meter) Propagate through op layer3.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer3.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 24]].
(nn-Meter) Output shape of layer3.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 24]].
(nn-Meter) Find node layer4.1.conv/Conv2D with its weight op layer4.1.conv/weight.
(nn-Meter) Get input shape of layer4.1.conv/Conv2D from layer3.3.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 28, 28, 24].
(nn-Meter) Get weight shape of layer4.1.conv/Conv2D from ['layer4.1.conv/weight'], input shape:[1, 1, 24, 88].
(nn-Meter) Op:layer4.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 28, 28, 24], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer4.1.conv/Conv2D op is [[1, 28, 28, 24]].
(nn-Meter) Output shape of layer4.1.conv/Conv2D op is [[1, 28, 28, 88]].
(nn-Meter) Propagate through op layer4.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer4.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 88]].
(nn-Meter) Output shape of layer4.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 88]].
(nn-Meter) Propagate through op layer4.1.relu/Relu.
(nn-Meter) Input shape of layer4.1.relu/Relu op is [[1, 28, 28, 88]].
(nn-Meter) Output shape of layer4.1.relu/Relu op is [[1, 28, 28, 88]].
(nn-Meter) Find node layer4.2.depconv/depthwise with its weight op layer4.2.depconv/weight.
(nn-Meter) Get input shape of layer4.2.depconv/depthwise from layer4.1.relu/Relu, input shape:[1, 28, 28, 88].
(nn-Meter) Get weight shape of layer4.2.depconv/depthwise from ['layer4.2.depconv/weight'], input shape:[3, 3, 88, 1].
(nn-Meter) Op:layer4.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 28, 28, 88], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer4.2.depconv/depthwise op is [[1, 28, 28, 88]].
(nn-Meter) Output shape of layer4.2.depconv/depthwise op is [[1, 28, 28, 88]].
(nn-Meter) Propagate through op layer4.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer4.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 88]].
(nn-Meter) Output shape of layer4.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 88]].
(nn-Meter) Propagate through op layer4.2.relu/Relu.
(nn-Meter) Input shape of layer4.2.relu/Relu op is [[1, 28, 28, 88]].
(nn-Meter) Output shape of layer4.2.relu/Relu op is [[1, 28, 28, 88]].
(nn-Meter) Find node layer4.3.conv/Conv2D with its weight op layer4.3.conv/weight.
(nn-Meter) Get input shape of layer4.3.conv/Conv2D from layer4.2.relu/Relu, input shape:[1, 28, 28, 88].
(nn-Meter) Get weight shape of layer4.3.conv/Conv2D from ['layer4.3.conv/weight'], input shape:[1, 1, 88, 24].
(nn-Meter) Op:layer4.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 28, 28, 88], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer4.3.conv/Conv2D op is [[1, 28, 28, 88]].
(nn-Meter) Output shape of layer4.3.conv/Conv2D op is [[1, 28, 28, 24]].
(nn-Meter) Propagate through op layer4.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer4.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 24]].
(nn-Meter) Output shape of layer4.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 24]].
(nn-Meter) Input shape of Add_1 op is [[1, 28, 28, 24], [1, 28, 28, 24]].
(nn-Meter) Output shape of Add_1 op is [[1, 28, 28, 24]].
(nn-Meter) Find node layer5.1.conv/Conv2D with its weight op layer5.1.conv/weight.
(nn-Meter) Get input shape of layer5.1.conv/Conv2D from Add_1, input shape:[1, 28, 28, 24].
(nn-Meter) Get weight shape of layer5.1.conv/Conv2D from ['layer5.1.conv/weight'], input shape:[1, 1, 24, 96].
(nn-Meter) Op:layer5.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 28, 28, 24], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer5.1.conv/Conv2D op is [[1, 28, 28, 24]].
(nn-Meter) Output shape of layer5.1.conv/Conv2D op is [[1, 28, 28, 96]].
(nn-Meter) Propagate through op layer5.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer5.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 96]].
(nn-Meter) Output shape of layer5.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 28, 28, 96]].
(nn-Meter) Input shape of layer5.1.hswish/add op is [[1, 28, 28, 96], []].
(nn-Meter) Output shape of layer5.1.hswish/add op is [[1, 28, 28, 96]].
(nn-Meter) Propagate through op layer5.1.hswish/Relu6.
(nn-Meter) Input shape of layer5.1.hswish/Relu6 op is [[1, 28, 28, 96]].
(nn-Meter) Output shape of layer5.1.hswish/Relu6 op is [[1, 28, 28, 96]].
(nn-Meter) Input shape of layer5.1.hswish/mul op is [[1, 28, 28, 96], []].
(nn-Meter) Output shape of layer5.1.hswish/mul op is [[1, 28, 28, 96]].
(nn-Meter) Input shape of layer5.1.hswish/mul_1 op is [[1, 28, 28, 96], [1, 28, 28, 96]].
(nn-Meter) Output shape of layer5.1.hswish/mul_1 op is [[1, 28, 28, 96]].
(nn-Meter) Find node layer5.2.depconv/depthwise with its weight op layer5.2.depconv/weight.
(nn-Meter) Get input shape of layer5.2.depconv/depthwise from layer5.1.hswish/mul_1, input shape:[1, 28, 28, 96].
(nn-Meter) Get weight shape of layer5.2.depconv/depthwise from ['layer5.2.depconv/weight'], input shape:[5, 5, 96, 1].
(nn-Meter) Op:layer5.2.depconv/depthwise, stride:[1, 2, 2, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 28, 28, 96], kernel size: [2, 2], strides: [1, 2, 2, 1], padding: SAME.
(nn-Meter) Input shape of layer5.2.depconv/depthwise op is [[1, 28, 28, 96]].
(nn-Meter) Output shape of layer5.2.depconv/depthwise op is [[1, 14, 14, 96]].
(nn-Meter) Propagate through op layer5.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer5.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 96]].
(nn-Meter) Output shape of layer5.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 96]].
(nn-Meter) Get input shape of SE_1/AvgPool from layer5.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 96].
(nn-Meter) Op:SE_1/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 96], kernel size: [1, 14, 14, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_1/AvgPool op is [[1, 14, 14, 96]].
(nn-Meter) Output shape of SE_1/AvgPool op is [[1, 1, 1, 96]].
(nn-Meter) Find node SE_1/conv2d_2/Conv2D with its weight op SE_1/conv2d_2/kernel.
(nn-Meter) Get input shape of SE_1/conv2d_2/Conv2D from SE_1/AvgPool, input shape:[1, 1, 1, 96].
(nn-Meter) Get weight shape of SE_1/conv2d_2/Conv2D from ['SE_1/conv2d_2/kernel'], input shape:[1, 1, 96, 24].
(nn-Meter) Op:SE_1/conv2d_2/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 96], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_1/conv2d_2/Conv2D op is [[1, 1, 1, 96]].
(nn-Meter) Output shape of SE_1/conv2d_2/Conv2D op is [[1, 1, 1, 24]].
(nn-Meter) Propagate through op SE_1/conv2d_2/BiasAdd.
(nn-Meter) Input shape of SE_1/conv2d_2/BiasAdd op is [[1, 1, 1, 24]].
(nn-Meter) Output shape of SE_1/conv2d_2/BiasAdd op is [[1, 1, 1, 24]].
(nn-Meter) Propagate through op SE_1/Relu.
(nn-Meter) Input shape of SE_1/Relu op is [[1, 1, 1, 24]].
(nn-Meter) Output shape of SE_1/Relu op is [[1, 1, 1, 24]].
(nn-Meter) Find node SE_1/conv2d_3/Conv2D with its weight op SE_1/conv2d_3/kernel.
(nn-Meter) Get input shape of SE_1/conv2d_3/Conv2D from SE_1/Relu, input shape:[1, 1, 1, 24].
(nn-Meter) Get weight shape of SE_1/conv2d_3/Conv2D from ['SE_1/conv2d_3/kernel'], input shape:[1, 1, 24, 96].
(nn-Meter) Op:SE_1/conv2d_3/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 24], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_1/conv2d_3/Conv2D op is [[1, 1, 1, 24]].
(nn-Meter) Output shape of SE_1/conv2d_3/Conv2D op is [[1, 1, 1, 96]].
(nn-Meter) Propagate through op SE_1/conv2d_3/BiasAdd.
(nn-Meter) Input shape of SE_1/conv2d_3/BiasAdd op is [[1, 1, 1, 96]].
(nn-Meter) Output shape of SE_1/conv2d_3/BiasAdd op is [[1, 1, 1, 96]].
(nn-Meter) Input shape of Add_2 op is [[1, 1, 1, 96], []].
(nn-Meter) Output shape of Add_2 op is [[1, 1, 1, 96]].
(nn-Meter) Propagate through op Relu6_1.
(nn-Meter) Input shape of Relu6_1 op is [[1, 1, 1, 96]].
(nn-Meter) Output shape of Relu6_1 op is [[1, 1, 1, 96]].
(nn-Meter) Input shape of mul_2 op is [[1, 1, 1, 96], []].
(nn-Meter) Output shape of mul_2 op is [[1, 1, 1, 96]].
(nn-Meter) Input shape of mul_3 op is [[1, 14, 14, 96], [1, 14, 14, 96]].
(nn-Meter) Output shape of mul_3 op is [[1, 14, 14, 96]].
(nn-Meter) Input shape of layer5.2.hswish/add op is [[1, 14, 14, 96], []].
(nn-Meter) Output shape of layer5.2.hswish/add op is [[1, 14, 14, 96]].
(nn-Meter) Propagate through op layer5.2.hswish/Relu6.
(nn-Meter) Input shape of layer5.2.hswish/Relu6 op is [[1, 14, 14, 96]].
(nn-Meter) Output shape of layer5.2.hswish/Relu6 op is [[1, 14, 14, 96]].
(nn-Meter) Input shape of layer5.2.hswish/mul op is [[1, 14, 14, 96], []].
(nn-Meter) Output shape of layer5.2.hswish/mul op is [[1, 14, 14, 96]].
(nn-Meter) Input shape of layer5.2.hswish/mul_1 op is [[1, 14, 14, 96], [1, 14, 14, 96]].
(nn-Meter) Output shape of layer5.2.hswish/mul_1 op is [[1, 14, 14, 96]].
(nn-Meter) Find node layer5.3.conv/Conv2D with its weight op layer5.3.conv/weight.
(nn-Meter) Get input shape of layer5.3.conv/Conv2D from layer5.2.hswish/mul_1, input shape:[1, 14, 14, 96].
(nn-Meter) Get weight shape of layer5.3.conv/Conv2D from ['layer5.3.conv/weight'], input shape:[1, 1, 96, 40].
(nn-Meter) Op:layer5.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 96], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer5.3.conv/Conv2D op is [[1, 14, 14, 96]].
(nn-Meter) Output shape of layer5.3.conv/Conv2D op is [[1, 14, 14, 40]].
(nn-Meter) Propagate through op layer5.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer5.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 40]].
(nn-Meter) Output shape of layer5.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 40]].
(nn-Meter) Find node layer6.1.conv/Conv2D with its weight op layer6.1.conv/weight.
(nn-Meter) Get input shape of layer6.1.conv/Conv2D from layer5.3.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 40].
(nn-Meter) Get weight shape of layer6.1.conv/Conv2D from ['layer6.1.conv/weight'], input shape:[1, 1, 40, 240].
(nn-Meter) Op:layer6.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 40], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer6.1.conv/Conv2D op is [[1, 14, 14, 40]].
(nn-Meter) Output shape of layer6.1.conv/Conv2D op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer6.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer6.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer6.1.hswish/add op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer6.1.hswish/add op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer6.1.hswish/Relu6.
(nn-Meter) Input shape of layer6.1.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.1.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer6.1.hswish/mul op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer6.1.hswish/mul op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer6.1.hswish/mul_1 op is [[1, 14, 14, 240], [1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.1.hswish/mul_1 op is [[1, 14, 14, 240]].
(nn-Meter) Find node layer6.2.depconv/depthwise with its weight op layer6.2.depconv/weight.
(nn-Meter) Get input shape of layer6.2.depconv/depthwise from layer6.1.hswish/mul_1, input shape:[1, 14, 14, 240].
(nn-Meter) Get weight shape of layer6.2.depconv/depthwise from ['layer6.2.depconv/weight'], input shape:[5, 5, 240, 1].
(nn-Meter) Op:layer6.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 240], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer6.2.depconv/depthwise op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.2.depconv/depthwise op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer6.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer6.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Get input shape of SE_2/AvgPool from layer6.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 240].
(nn-Meter) Op:SE_2/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 240], kernel size: [1, 14, 14, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_2/AvgPool op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of SE_2/AvgPool op is [[1, 1, 1, 240]].
(nn-Meter) Find node SE_2/conv2d_4/Conv2D with its weight op SE_2/conv2d_4/kernel.
(nn-Meter) Get input shape of SE_2/conv2d_4/Conv2D from SE_2/AvgPool, input shape:[1, 1, 1, 240].
(nn-Meter) Get weight shape of SE_2/conv2d_4/Conv2D from ['SE_2/conv2d_4/kernel'], input shape:[1, 1, 240, 60].
(nn-Meter) Op:SE_2/conv2d_4/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 240], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_2/conv2d_4/Conv2D op is [[1, 1, 1, 240]].
(nn-Meter) Output shape of SE_2/conv2d_4/Conv2D op is [[1, 1, 1, 60]].
(nn-Meter) Propagate through op SE_2/conv2d_4/BiasAdd.
(nn-Meter) Input shape of SE_2/conv2d_4/BiasAdd op is [[1, 1, 1, 60]].
(nn-Meter) Output shape of SE_2/conv2d_4/BiasAdd op is [[1, 1, 1, 60]].
(nn-Meter) Propagate through op SE_2/Relu.
(nn-Meter) Input shape of SE_2/Relu op is [[1, 1, 1, 60]].
(nn-Meter) Output shape of SE_2/Relu op is [[1, 1, 1, 60]].
(nn-Meter) Find node SE_2/conv2d_5/Conv2D with its weight op SE_2/conv2d_5/kernel.
(nn-Meter) Get input shape of SE_2/conv2d_5/Conv2D from SE_2/Relu, input shape:[1, 1, 1, 60].
(nn-Meter) Get weight shape of SE_2/conv2d_5/Conv2D from ['SE_2/conv2d_5/kernel'], input shape:[1, 1, 60, 240].
(nn-Meter) Op:SE_2/conv2d_5/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 60], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_2/conv2d_5/Conv2D op is [[1, 1, 1, 60]].
(nn-Meter) Output shape of SE_2/conv2d_5/Conv2D op is [[1, 1, 1, 240]].
(nn-Meter) Propagate through op SE_2/conv2d_5/BiasAdd.
(nn-Meter) Input shape of SE_2/conv2d_5/BiasAdd op is [[1, 1, 1, 240]].
(nn-Meter) Output shape of SE_2/conv2d_5/BiasAdd op is [[1, 1, 1, 240]].
(nn-Meter) Input shape of Add_3 op is [[1, 1, 1, 240], []].
(nn-Meter) Output shape of Add_3 op is [[1, 1, 1, 240]].
(nn-Meter) Propagate through op Relu6_2.
(nn-Meter) Input shape of Relu6_2 op is [[1, 1, 1, 240]].
(nn-Meter) Output shape of Relu6_2 op is [[1, 1, 1, 240]].
(nn-Meter) Input shape of mul_4 op is [[1, 1, 1, 240], []].
(nn-Meter) Output shape of mul_4 op is [[1, 1, 1, 240]].
(nn-Meter) Input shape of mul_5 op is [[1, 14, 14, 240], [1, 14, 14, 240]].
(nn-Meter) Output shape of mul_5 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer6.2.hswish/add op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer6.2.hswish/add op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer6.2.hswish/Relu6.
(nn-Meter) Input shape of layer6.2.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.2.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer6.2.hswish/mul op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer6.2.hswish/mul op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer6.2.hswish/mul_1 op is [[1, 14, 14, 240], [1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.2.hswish/mul_1 op is [[1, 14, 14, 240]].
(nn-Meter) Find node layer6.3.conv/Conv2D with its weight op layer6.3.conv/weight.
(nn-Meter) Get input shape of layer6.3.conv/Conv2D from layer6.2.hswish/mul_1, input shape:[1, 14, 14, 240].
(nn-Meter) Get weight shape of layer6.3.conv/Conv2D from ['layer6.3.conv/weight'], input shape:[1, 1, 240, 40].
(nn-Meter) Op:layer6.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 240], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer6.3.conv/Conv2D op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer6.3.conv/Conv2D op is [[1, 14, 14, 40]].
(nn-Meter) Propagate through op layer6.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer6.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 40]].
(nn-Meter) Output shape of layer6.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 40]].
(nn-Meter) Input shape of Add_4 op is [[1, 14, 14, 40], [1, 14, 14, 40]].
(nn-Meter) Output shape of Add_4 op is [[1, 14, 14, 40]].
(nn-Meter) Find node layer7.1.conv/Conv2D with its weight op layer7.1.conv/weight.
(nn-Meter) Get input shape of layer7.1.conv/Conv2D from Add_4, input shape:[1, 14, 14, 40].
(nn-Meter) Get weight shape of layer7.1.conv/Conv2D from ['layer7.1.conv/weight'], input shape:[1, 1, 40, 240].
(nn-Meter) Op:layer7.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 40], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer7.1.conv/Conv2D op is [[1, 14, 14, 40]].
(nn-Meter) Output shape of layer7.1.conv/Conv2D op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer7.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer7.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer7.1.hswish/add op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer7.1.hswish/add op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer7.1.hswish/Relu6.
(nn-Meter) Input shape of layer7.1.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.1.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer7.1.hswish/mul op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer7.1.hswish/mul op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer7.1.hswish/mul_1 op is [[1, 14, 14, 240], [1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.1.hswish/mul_1 op is [[1, 14, 14, 240]].
(nn-Meter) Find node layer7.2.depconv/depthwise with its weight op layer7.2.depconv/weight.
(nn-Meter) Get input shape of layer7.2.depconv/depthwise from layer7.1.hswish/mul_1, input shape:[1, 14, 14, 240].
(nn-Meter) Get weight shape of layer7.2.depconv/depthwise from ['layer7.2.depconv/weight'], input shape:[5, 5, 240, 1].
(nn-Meter) Op:layer7.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 240], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer7.2.depconv/depthwise op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.2.depconv/depthwise op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer7.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer7.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 240]].
(nn-Meter) Get input shape of SE_3/AvgPool from layer7.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 240].
(nn-Meter) Op:SE_3/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 240], kernel size: [1, 14, 14, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_3/AvgPool op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of SE_3/AvgPool op is [[1, 1, 1, 240]].
(nn-Meter) Find node SE_3/conv2d_6/Conv2D with its weight op SE_3/conv2d_6/kernel.
(nn-Meter) Get input shape of SE_3/conv2d_6/Conv2D from SE_3/AvgPool, input shape:[1, 1, 1, 240].
(nn-Meter) Get weight shape of SE_3/conv2d_6/Conv2D from ['SE_3/conv2d_6/kernel'], input shape:[1, 1, 240, 60].
(nn-Meter) Op:SE_3/conv2d_6/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 240], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_3/conv2d_6/Conv2D op is [[1, 1, 1, 240]].
(nn-Meter) Output shape of SE_3/conv2d_6/Conv2D op is [[1, 1, 1, 60]].
(nn-Meter) Propagate through op SE_3/conv2d_6/BiasAdd.
(nn-Meter) Input shape of SE_3/conv2d_6/BiasAdd op is [[1, 1, 1, 60]].
(nn-Meter) Output shape of SE_3/conv2d_6/BiasAdd op is [[1, 1, 1, 60]].
(nn-Meter) Propagate through op SE_3/Relu.
(nn-Meter) Input shape of SE_3/Relu op is [[1, 1, 1, 60]].
(nn-Meter) Output shape of SE_3/Relu op is [[1, 1, 1, 60]].
(nn-Meter) Find node SE_3/conv2d_7/Conv2D with its weight op SE_3/conv2d_7/kernel.
(nn-Meter) Get input shape of SE_3/conv2d_7/Conv2D from SE_3/Relu, input shape:[1, 1, 1, 60].
(nn-Meter) Get weight shape of SE_3/conv2d_7/Conv2D from ['SE_3/conv2d_7/kernel'], input shape:[1, 1, 60, 240].
(nn-Meter) Op:SE_3/conv2d_7/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 60], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_3/conv2d_7/Conv2D op is [[1, 1, 1, 60]].
(nn-Meter) Output shape of SE_3/conv2d_7/Conv2D op is [[1, 1, 1, 240]].
(nn-Meter) Propagate through op SE_3/conv2d_7/BiasAdd.
(nn-Meter) Input shape of SE_3/conv2d_7/BiasAdd op is [[1, 1, 1, 240]].
(nn-Meter) Output shape of SE_3/conv2d_7/BiasAdd op is [[1, 1, 1, 240]].
(nn-Meter) Input shape of Add_5 op is [[1, 1, 1, 240], []].
(nn-Meter) Output shape of Add_5 op is [[1, 1, 1, 240]].
(nn-Meter) Propagate through op Relu6_3.
(nn-Meter) Input shape of Relu6_3 op is [[1, 1, 1, 240]].
(nn-Meter) Output shape of Relu6_3 op is [[1, 1, 1, 240]].
(nn-Meter) Input shape of mul_6 op is [[1, 1, 1, 240], []].
(nn-Meter) Output shape of mul_6 op is [[1, 1, 1, 240]].
(nn-Meter) Input shape of mul_7 op is [[1, 14, 14, 240], [1, 14, 14, 240]].
(nn-Meter) Output shape of mul_7 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer7.2.hswish/add op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer7.2.hswish/add op is [[1, 14, 14, 240]].
(nn-Meter) Propagate through op layer7.2.hswish/Relu6.
(nn-Meter) Input shape of layer7.2.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.2.hswish/Relu6 op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer7.2.hswish/mul op is [[1, 14, 14, 240], []].
(nn-Meter) Output shape of layer7.2.hswish/mul op is [[1, 14, 14, 240]].
(nn-Meter) Input shape of layer7.2.hswish/mul_1 op is [[1, 14, 14, 240], [1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.2.hswish/mul_1 op is [[1, 14, 14, 240]].
(nn-Meter) Find node layer7.3.conv/Conv2D with its weight op layer7.3.conv/weight.
(nn-Meter) Get input shape of layer7.3.conv/Conv2D from layer7.2.hswish/mul_1, input shape:[1, 14, 14, 240].
(nn-Meter) Get weight shape of layer7.3.conv/Conv2D from ['layer7.3.conv/weight'], input shape:[1, 1, 240, 40].
(nn-Meter) Op:layer7.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 240], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer7.3.conv/Conv2D op is [[1, 14, 14, 240]].
(nn-Meter) Output shape of layer7.3.conv/Conv2D op is [[1, 14, 14, 40]].
(nn-Meter) Propagate through op layer7.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer7.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 40]].
(nn-Meter) Output shape of layer7.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 40]].
(nn-Meter) Input shape of Add_6 op is [[1, 14, 14, 40], [1, 14, 14, 40]].
(nn-Meter) Output shape of Add_6 op is [[1, 14, 14, 40]].
(nn-Meter) Find node layer8.1.conv/Conv2D with its weight op layer8.1.conv/weight.
(nn-Meter) Get input shape of layer8.1.conv/Conv2D from Add_6, input shape:[1, 14, 14, 40].
(nn-Meter) Get weight shape of layer8.1.conv/Conv2D from ['layer8.1.conv/weight'], input shape:[1, 1, 40, 120].
(nn-Meter) Op:layer8.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 40], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer8.1.conv/Conv2D op is [[1, 14, 14, 40]].
(nn-Meter) Output shape of layer8.1.conv/Conv2D op is [[1, 14, 14, 120]].
(nn-Meter) Propagate through op layer8.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer8.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 120]].
(nn-Meter) Input shape of layer8.1.hswish/add op is [[1, 14, 14, 120], []].
(nn-Meter) Output shape of layer8.1.hswish/add op is [[1, 14, 14, 120]].
(nn-Meter) Propagate through op layer8.1.hswish/Relu6.
(nn-Meter) Input shape of layer8.1.hswish/Relu6 op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.1.hswish/Relu6 op is [[1, 14, 14, 120]].
(nn-Meter) Input shape of layer8.1.hswish/mul op is [[1, 14, 14, 120], []].
(nn-Meter) Output shape of layer8.1.hswish/mul op is [[1, 14, 14, 120]].
(nn-Meter) Input shape of layer8.1.hswish/mul_1 op is [[1, 14, 14, 120], [1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.1.hswish/mul_1 op is [[1, 14, 14, 120]].
(nn-Meter) Find node layer8.2.depconv/depthwise with its weight op layer8.2.depconv/weight.
(nn-Meter) Get input shape of layer8.2.depconv/depthwise from layer8.1.hswish/mul_1, input shape:[1, 14, 14, 120].
(nn-Meter) Get weight shape of layer8.2.depconv/depthwise from ['layer8.2.depconv/weight'], input shape:[5, 5, 120, 1].
(nn-Meter) Op:layer8.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 120], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer8.2.depconv/depthwise op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.2.depconv/depthwise op is [[1, 14, 14, 120]].
(nn-Meter) Propagate through op layer8.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer8.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 120]].
(nn-Meter) Get input shape of SE_4/AvgPool from layer8.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 120].
(nn-Meter) Op:SE_4/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 120], kernel size: [1, 14, 14, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_4/AvgPool op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of SE_4/AvgPool op is [[1, 1, 1, 120]].
(nn-Meter) Find node SE_4/conv2d_8/Conv2D with its weight op SE_4/conv2d_8/kernel.
(nn-Meter) Get input shape of SE_4/conv2d_8/Conv2D from SE_4/AvgPool, input shape:[1, 1, 1, 120].
(nn-Meter) Get weight shape of SE_4/conv2d_8/Conv2D from ['SE_4/conv2d_8/kernel'], input shape:[1, 1, 120, 30].
(nn-Meter) Op:SE_4/conv2d_8/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 120], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_4/conv2d_8/Conv2D op is [[1, 1, 1, 120]].
(nn-Meter) Output shape of SE_4/conv2d_8/Conv2D op is [[1, 1, 1, 30]].
(nn-Meter) Propagate through op SE_4/conv2d_8/BiasAdd.
(nn-Meter) Input shape of SE_4/conv2d_8/BiasAdd op is [[1, 1, 1, 30]].
(nn-Meter) Output shape of SE_4/conv2d_8/BiasAdd op is [[1, 1, 1, 30]].
(nn-Meter) Propagate through op SE_4/Relu.
(nn-Meter) Input shape of SE_4/Relu op is [[1, 1, 1, 30]].
(nn-Meter) Output shape of SE_4/Relu op is [[1, 1, 1, 30]].
(nn-Meter) Find node SE_4/conv2d_9/Conv2D with its weight op SE_4/conv2d_9/kernel.
(nn-Meter) Get input shape of SE_4/conv2d_9/Conv2D from SE_4/Relu, input shape:[1, 1, 1, 30].
(nn-Meter) Get weight shape of SE_4/conv2d_9/Conv2D from ['SE_4/conv2d_9/kernel'], input shape:[1, 1, 30, 120].
(nn-Meter) Op:SE_4/conv2d_9/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 30], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_4/conv2d_9/Conv2D op is [[1, 1, 1, 30]].
(nn-Meter) Output shape of SE_4/conv2d_9/Conv2D op is [[1, 1, 1, 120]].
(nn-Meter) Propagate through op SE_4/conv2d_9/BiasAdd.
(nn-Meter) Input shape of SE_4/conv2d_9/BiasAdd op is [[1, 1, 1, 120]].
(nn-Meter) Output shape of SE_4/conv2d_9/BiasAdd op is [[1, 1, 1, 120]].
(nn-Meter) Input shape of Add_7 op is [[1, 1, 1, 120], []].
(nn-Meter) Output shape of Add_7 op is [[1, 1, 1, 120]].
(nn-Meter) Propagate through op Relu6_4.
(nn-Meter) Input shape of Relu6_4 op is [[1, 1, 1, 120]].
(nn-Meter) Output shape of Relu6_4 op is [[1, 1, 1, 120]].
(nn-Meter) Input shape of mul_8 op is [[1, 1, 1, 120], []].
(nn-Meter) Output shape of mul_8 op is [[1, 1, 1, 120]].
(nn-Meter) Input shape of mul_9 op is [[1, 14, 14, 120], [1, 14, 14, 120]].
(nn-Meter) Output shape of mul_9 op is [[1, 14, 14, 120]].
(nn-Meter) Input shape of layer8.2.hswish/add op is [[1, 14, 14, 120], []].
(nn-Meter) Output shape of layer8.2.hswish/add op is [[1, 14, 14, 120]].
(nn-Meter) Propagate through op layer8.2.hswish/Relu6.
(nn-Meter) Input shape of layer8.2.hswish/Relu6 op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.2.hswish/Relu6 op is [[1, 14, 14, 120]].
(nn-Meter) Input shape of layer8.2.hswish/mul op is [[1, 14, 14, 120], []].
(nn-Meter) Output shape of layer8.2.hswish/mul op is [[1, 14, 14, 120]].
(nn-Meter) Input shape of layer8.2.hswish/mul_1 op is [[1, 14, 14, 120], [1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.2.hswish/mul_1 op is [[1, 14, 14, 120]].
(nn-Meter) Find node layer8.3.conv/Conv2D with its weight op layer8.3.conv/weight.
(nn-Meter) Get input shape of layer8.3.conv/Conv2D from layer8.2.hswish/mul_1, input shape:[1, 14, 14, 120].
(nn-Meter) Get weight shape of layer8.3.conv/Conv2D from ['layer8.3.conv/weight'], input shape:[1, 1, 120, 48].
(nn-Meter) Op:layer8.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 120], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer8.3.conv/Conv2D op is [[1, 14, 14, 120]].
(nn-Meter) Output shape of layer8.3.conv/Conv2D op is [[1, 14, 14, 48]].
(nn-Meter) Propagate through op layer8.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer8.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 48]].
(nn-Meter) Output shape of layer8.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 48]].
(nn-Meter) Find node layer9.1.conv/Conv2D with its weight op layer9.1.conv/weight.
(nn-Meter) Get input shape of layer9.1.conv/Conv2D from layer8.3.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 48].
(nn-Meter) Get weight shape of layer9.1.conv/Conv2D from ['layer9.1.conv/weight'], input shape:[1, 1, 48, 144].
(nn-Meter) Op:layer9.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 48], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer9.1.conv/Conv2D op is [[1, 14, 14, 48]].
(nn-Meter) Output shape of layer9.1.conv/Conv2D op is [[1, 14, 14, 144]].
(nn-Meter) Propagate through op layer9.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer9.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 144]].
(nn-Meter) Input shape of layer9.1.hswish/add op is [[1, 14, 14, 144], []].
(nn-Meter) Output shape of layer9.1.hswish/add op is [[1, 14, 14, 144]].
(nn-Meter) Propagate through op layer9.1.hswish/Relu6.
(nn-Meter) Input shape of layer9.1.hswish/Relu6 op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.1.hswish/Relu6 op is [[1, 14, 14, 144]].
(nn-Meter) Input shape of layer9.1.hswish/mul op is [[1, 14, 14, 144], []].
(nn-Meter) Output shape of layer9.1.hswish/mul op is [[1, 14, 14, 144]].
(nn-Meter) Input shape of layer9.1.hswish/mul_1 op is [[1, 14, 14, 144], [1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.1.hswish/mul_1 op is [[1, 14, 14, 144]].
(nn-Meter) Find node layer9.2.depconv/depthwise with its weight op layer9.2.depconv/weight.
(nn-Meter) Get input shape of layer9.2.depconv/depthwise from layer9.1.hswish/mul_1, input shape:[1, 14, 14, 144].
(nn-Meter) Get weight shape of layer9.2.depconv/depthwise from ['layer9.2.depconv/weight'], input shape:[5, 5, 144, 1].
(nn-Meter) Op:layer9.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 144], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer9.2.depconv/depthwise op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.2.depconv/depthwise op is [[1, 14, 14, 144]].
(nn-Meter) Propagate through op layer9.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer9.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 144]].
(nn-Meter) Get input shape of SE_5/AvgPool from layer9.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 14, 14, 144].
(nn-Meter) Op:SE_5/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 144], kernel size: [1, 14, 14, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_5/AvgPool op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of SE_5/AvgPool op is [[1, 1, 1, 144]].
(nn-Meter) Find node SE_5/conv2d_10/Conv2D with its weight op SE_5/conv2d_10/kernel.
(nn-Meter) Get input shape of SE_5/conv2d_10/Conv2D from SE_5/AvgPool, input shape:[1, 1, 1, 144].
(nn-Meter) Get weight shape of SE_5/conv2d_10/Conv2D from ['SE_5/conv2d_10/kernel'], input shape:[1, 1, 144, 36].
(nn-Meter) Op:SE_5/conv2d_10/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 144], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_5/conv2d_10/Conv2D op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_5/conv2d_10/Conv2D op is [[1, 1, 1, 36]].
(nn-Meter) Propagate through op SE_5/conv2d_10/BiasAdd.
(nn-Meter) Input shape of SE_5/conv2d_10/BiasAdd op is [[1, 1, 1, 36]].
(nn-Meter) Output shape of SE_5/conv2d_10/BiasAdd op is [[1, 1, 1, 36]].
(nn-Meter) Propagate through op SE_5/Relu.
(nn-Meter) Input shape of SE_5/Relu op is [[1, 1, 1, 36]].
(nn-Meter) Output shape of SE_5/Relu op is [[1, 1, 1, 36]].
(nn-Meter) Find node SE_5/conv2d_11/Conv2D with its weight op SE_5/conv2d_11/kernel.
(nn-Meter) Get input shape of SE_5/conv2d_11/Conv2D from SE_5/Relu, input shape:[1, 1, 1, 36].
(nn-Meter) Get weight shape of SE_5/conv2d_11/Conv2D from ['SE_5/conv2d_11/kernel'], input shape:[1, 1, 36, 144].
(nn-Meter) Op:SE_5/conv2d_11/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 36], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_5/conv2d_11/Conv2D op is [[1, 1, 1, 36]].
(nn-Meter) Output shape of SE_5/conv2d_11/Conv2D op is [[1, 1, 1, 144]].
(nn-Meter) Propagate through op SE_5/conv2d_11/BiasAdd.
(nn-Meter) Input shape of SE_5/conv2d_11/BiasAdd op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_5/conv2d_11/BiasAdd op is [[1, 1, 1, 144]].
(nn-Meter) Input shape of Add_8 op is [[1, 1, 1, 144], []].
(nn-Meter) Output shape of Add_8 op is [[1, 1, 1, 144]].
(nn-Meter) Propagate through op Relu6_5.
(nn-Meter) Input shape of Relu6_5 op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of Relu6_5 op is [[1, 1, 1, 144]].
(nn-Meter) Input shape of mul_10 op is [[1, 1, 1, 144], []].
(nn-Meter) Output shape of mul_10 op is [[1, 1, 1, 144]].
(nn-Meter) Input shape of mul_11 op is [[1, 14, 14, 144], [1, 14, 14, 144]].
(nn-Meter) Output shape of mul_11 op is [[1, 14, 14, 144]].
(nn-Meter) Input shape of layer9.2.hswish/add op is [[1, 14, 14, 144], []].
(nn-Meter) Output shape of layer9.2.hswish/add op is [[1, 14, 14, 144]].
(nn-Meter) Propagate through op layer9.2.hswish/Relu6.
(nn-Meter) Input shape of layer9.2.hswish/Relu6 op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.2.hswish/Relu6 op is [[1, 14, 14, 144]].
(nn-Meter) Input shape of layer9.2.hswish/mul op is [[1, 14, 14, 144], []].
(nn-Meter) Output shape of layer9.2.hswish/mul op is [[1, 14, 14, 144]].
(nn-Meter) Input shape of layer9.2.hswish/mul_1 op is [[1, 14, 14, 144], [1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.2.hswish/mul_1 op is [[1, 14, 14, 144]].
(nn-Meter) Find node layer9.3.conv/Conv2D with its weight op layer9.3.conv/weight.
(nn-Meter) Get input shape of layer9.3.conv/Conv2D from layer9.2.hswish/mul_1, input shape:[1, 14, 14, 144].
(nn-Meter) Get weight shape of layer9.3.conv/Conv2D from ['layer9.3.conv/weight'], input shape:[1, 1, 144, 48].
(nn-Meter) Op:layer9.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 144], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer9.3.conv/Conv2D op is [[1, 14, 14, 144]].
(nn-Meter) Output shape of layer9.3.conv/Conv2D op is [[1, 14, 14, 48]].
(nn-Meter) Propagate through op layer9.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer9.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 48]].
(nn-Meter) Output shape of layer9.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 48]].
(nn-Meter) Input shape of Add_9 op is [[1, 14, 14, 48], [1, 14, 14, 48]].
(nn-Meter) Output shape of Add_9 op is [[1, 14, 14, 48]].
(nn-Meter) Find node layer10.1.conv/Conv2D with its weight op layer10.1.conv/weight.
(nn-Meter) Get input shape of layer10.1.conv/Conv2D from Add_9, input shape:[1, 14, 14, 48].
(nn-Meter) Get weight shape of layer10.1.conv/Conv2D from ['layer10.1.conv/weight'], input shape:[1, 1, 48, 288].
(nn-Meter) Op:layer10.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 48], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer10.1.conv/Conv2D op is [[1, 14, 14, 48]].
(nn-Meter) Output shape of layer10.1.conv/Conv2D op is [[1, 14, 14, 288]].
(nn-Meter) Propagate through op layer10.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer10.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 288]].
(nn-Meter) Output shape of layer10.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 14, 14, 288]].
(nn-Meter) Input shape of layer10.1.hswish/add op is [[1, 14, 14, 288], []].
(nn-Meter) Output shape of layer10.1.hswish/add op is [[1, 14, 14, 288]].
(nn-Meter) Propagate through op layer10.1.hswish/Relu6.
(nn-Meter) Input shape of layer10.1.hswish/Relu6 op is [[1, 14, 14, 288]].
(nn-Meter) Output shape of layer10.1.hswish/Relu6 op is [[1, 14, 14, 288]].
(nn-Meter) Input shape of layer10.1.hswish/mul op is [[1, 14, 14, 288], []].
(nn-Meter) Output shape of layer10.1.hswish/mul op is [[1, 14, 14, 288]].
(nn-Meter) Input shape of layer10.1.hswish/mul_1 op is [[1, 14, 14, 288], [1, 14, 14, 288]].
(nn-Meter) Output shape of layer10.1.hswish/mul_1 op is [[1, 14, 14, 288]].
(nn-Meter) Find node layer10.2.depconv/depthwise with its weight op layer10.2.depconv/weight.
(nn-Meter) Get input shape of layer10.2.depconv/depthwise from layer10.1.hswish/mul_1, input shape:[1, 14, 14, 288].
(nn-Meter) Get weight shape of layer10.2.depconv/depthwise from ['layer10.2.depconv/weight'], input shape:[5, 5, 288, 1].
(nn-Meter) Op:layer10.2.depconv/depthwise, stride:[1, 2, 2, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 14, 14, 288], kernel size: [2, 2], strides: [1, 2, 2, 1], padding: SAME.
(nn-Meter) Input shape of layer10.2.depconv/depthwise op is [[1, 14, 14, 288]].
(nn-Meter) Output shape of layer10.2.depconv/depthwise op is [[1, 7, 7, 288]].
(nn-Meter) Propagate through op layer10.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer10.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 288]].
(nn-Meter) Output shape of layer10.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 288]].
(nn-Meter) Get input shape of SE_6/AvgPool from layer10.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 7, 7, 288].
(nn-Meter) Op:SE_6/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 288], kernel size: [1, 7, 7, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_6/AvgPool op is [[1, 7, 7, 288]].
(nn-Meter) Output shape of SE_6/AvgPool op is [[1, 1, 1, 288]].
(nn-Meter) Find node SE_6/conv2d_12/Conv2D with its weight op SE_6/conv2d_12/kernel.
(nn-Meter) Get input shape of SE_6/conv2d_12/Conv2D from SE_6/AvgPool, input shape:[1, 1, 1, 288].
(nn-Meter) Get weight shape of SE_6/conv2d_12/Conv2D from ['SE_6/conv2d_12/kernel'], input shape:[1, 1, 288, 72].
(nn-Meter) Op:SE_6/conv2d_12/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 288], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_6/conv2d_12/Conv2D op is [[1, 1, 1, 288]].
(nn-Meter) Output shape of SE_6/conv2d_12/Conv2D op is [[1, 1, 1, 72]].
(nn-Meter) Propagate through op SE_6/conv2d_12/BiasAdd.
(nn-Meter) Input shape of SE_6/conv2d_12/BiasAdd op is [[1, 1, 1, 72]].
(nn-Meter) Output shape of SE_6/conv2d_12/BiasAdd op is [[1, 1, 1, 72]].
(nn-Meter) Propagate through op SE_6/Relu.
(nn-Meter) Input shape of SE_6/Relu op is [[1, 1, 1, 72]].
(nn-Meter) Output shape of SE_6/Relu op is [[1, 1, 1, 72]].
(nn-Meter) Find node SE_6/conv2d_13/Conv2D with its weight op SE_6/conv2d_13/kernel.
(nn-Meter) Get input shape of SE_6/conv2d_13/Conv2D from SE_6/Relu, input shape:[1, 1, 1, 72].
(nn-Meter) Get weight shape of SE_6/conv2d_13/Conv2D from ['SE_6/conv2d_13/kernel'], input shape:[1, 1, 72, 288].
(nn-Meter) Op:SE_6/conv2d_13/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 72], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_6/conv2d_13/Conv2D op is [[1, 1, 1, 72]].
(nn-Meter) Output shape of SE_6/conv2d_13/Conv2D op is [[1, 1, 1, 288]].
(nn-Meter) Propagate through op SE_6/conv2d_13/BiasAdd.
(nn-Meter) Input shape of SE_6/conv2d_13/BiasAdd op is [[1, 1, 1, 288]].
(nn-Meter) Output shape of SE_6/conv2d_13/BiasAdd op is [[1, 1, 1, 288]].
(nn-Meter) Input shape of Add_10 op is [[1, 1, 1, 288], []].
(nn-Meter) Output shape of Add_10 op is [[1, 1, 1, 288]].
(nn-Meter) Propagate through op Relu6_6.
(nn-Meter) Input shape of Relu6_6 op is [[1, 1, 1, 288]].
(nn-Meter) Output shape of Relu6_6 op is [[1, 1, 1, 288]].
(nn-Meter) Input shape of mul_12 op is [[1, 1, 1, 288], []].
(nn-Meter) Output shape of mul_12 op is [[1, 1, 1, 288]].
(nn-Meter) Input shape of mul_13 op is [[1, 7, 7, 288], [1, 7, 7, 288]].
(nn-Meter) Output shape of mul_13 op is [[1, 7, 7, 288]].
(nn-Meter) Input shape of layer10.2.hswish/add op is [[1, 7, 7, 288], []].
(nn-Meter) Output shape of layer10.2.hswish/add op is [[1, 7, 7, 288]].
(nn-Meter) Propagate through op layer10.2.hswish/Relu6.
(nn-Meter) Input shape of layer10.2.hswish/Relu6 op is [[1, 7, 7, 288]].
(nn-Meter) Output shape of layer10.2.hswish/Relu6 op is [[1, 7, 7, 288]].
(nn-Meter) Input shape of layer10.2.hswish/mul op is [[1, 7, 7, 288], []].
(nn-Meter) Output shape of layer10.2.hswish/mul op is [[1, 7, 7, 288]].
(nn-Meter) Input shape of layer10.2.hswish/mul_1 op is [[1, 7, 7, 288], [1, 7, 7, 288]].
(nn-Meter) Output shape of layer10.2.hswish/mul_1 op is [[1, 7, 7, 288]].
(nn-Meter) Find node layer10.3.conv/Conv2D with its weight op layer10.3.conv/weight.
(nn-Meter) Get input shape of layer10.3.conv/Conv2D from layer10.2.hswish/mul_1, input shape:[1, 7, 7, 288].
(nn-Meter) Get weight shape of layer10.3.conv/Conv2D from ['layer10.3.conv/weight'], input shape:[1, 1, 288, 96].
(nn-Meter) Op:layer10.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 288], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer10.3.conv/Conv2D op is [[1, 7, 7, 288]].
(nn-Meter) Output shape of layer10.3.conv/Conv2D op is [[1, 7, 7, 96]].
(nn-Meter) Propagate through op layer10.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer10.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 96]].
(nn-Meter) Output shape of layer10.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 96]].
(nn-Meter) Find node layer11.1.conv/Conv2D with its weight op layer11.1.conv/weight.
(nn-Meter) Get input shape of layer11.1.conv/Conv2D from layer10.3.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 7, 7, 96].
(nn-Meter) Get weight shape of layer11.1.conv/Conv2D from ['layer11.1.conv/weight'], input shape:[1, 1, 96, 576].
(nn-Meter) Op:layer11.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 96], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer11.1.conv/Conv2D op is [[1, 7, 7, 96]].
(nn-Meter) Output shape of layer11.1.conv/Conv2D op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer11.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer11.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer11.1.hswish/add op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer11.1.hswish/add op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer11.1.hswish/Relu6.
(nn-Meter) Input shape of layer11.1.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.1.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer11.1.hswish/mul op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer11.1.hswish/mul op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer11.1.hswish/mul_1 op is [[1, 7, 7, 576], [1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.1.hswish/mul_1 op is [[1, 7, 7, 576]].
(nn-Meter) Find node layer11.2.depconv/depthwise with its weight op layer11.2.depconv/weight.
(nn-Meter) Get input shape of layer11.2.depconv/depthwise from layer11.1.hswish/mul_1, input shape:[1, 7, 7, 576].
(nn-Meter) Get weight shape of layer11.2.depconv/depthwise from ['layer11.2.depconv/weight'], input shape:[5, 5, 576, 1].
(nn-Meter) Op:layer11.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 576], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer11.2.depconv/depthwise op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.2.depconv/depthwise op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer11.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer11.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Get input shape of SE_7/AvgPool from layer11.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 7, 7, 576].
(nn-Meter) Op:SE_7/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 576], kernel size: [1, 7, 7, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_7/AvgPool op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of SE_7/AvgPool op is [[1, 1, 1, 576]].
(nn-Meter) Find node SE_7/conv2d_14/Conv2D with its weight op SE_7/conv2d_14/kernel.
(nn-Meter) Get input shape of SE_7/conv2d_14/Conv2D from SE_7/AvgPool, input shape:[1, 1, 1, 576].
(nn-Meter) Get weight shape of SE_7/conv2d_14/Conv2D from ['SE_7/conv2d_14/kernel'], input shape:[1, 1, 576, 144].
(nn-Meter) Op:SE_7/conv2d_14/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 576], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_7/conv2d_14/Conv2D op is [[1, 1, 1, 576]].
(nn-Meter) Output shape of SE_7/conv2d_14/Conv2D op is [[1, 1, 1, 144]].
(nn-Meter) Propagate through op SE_7/conv2d_14/BiasAdd.
(nn-Meter) Input shape of SE_7/conv2d_14/BiasAdd op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_7/conv2d_14/BiasAdd op is [[1, 1, 1, 144]].
(nn-Meter) Propagate through op SE_7/Relu.
(nn-Meter) Input shape of SE_7/Relu op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_7/Relu op is [[1, 1, 1, 144]].
(nn-Meter) Find node SE_7/conv2d_15/Conv2D with its weight op SE_7/conv2d_15/kernel.
(nn-Meter) Get input shape of SE_7/conv2d_15/Conv2D from SE_7/Relu, input shape:[1, 1, 1, 144].
(nn-Meter) Get weight shape of SE_7/conv2d_15/Conv2D from ['SE_7/conv2d_15/kernel'], input shape:[1, 1, 144, 576].
(nn-Meter) Op:SE_7/conv2d_15/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 144], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_7/conv2d_15/Conv2D op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_7/conv2d_15/Conv2D op is [[1, 1, 1, 576]].
(nn-Meter) Propagate through op SE_7/conv2d_15/BiasAdd.
(nn-Meter) Input shape of SE_7/conv2d_15/BiasAdd op is [[1, 1, 1, 576]].
(nn-Meter) Output shape of SE_7/conv2d_15/BiasAdd op is [[1, 1, 1, 576]].
(nn-Meter) Input shape of Add_11 op is [[1, 1, 1, 576], []].
(nn-Meter) Output shape of Add_11 op is [[1, 1, 1, 576]].
(nn-Meter) Propagate through op Relu6_7.
(nn-Meter) Input shape of Relu6_7 op is [[1, 1, 1, 576]].
(nn-Meter) Output shape of Relu6_7 op is [[1, 1, 1, 576]].
(nn-Meter) Input shape of mul_14 op is [[1, 1, 1, 576], []].
(nn-Meter) Output shape of mul_14 op is [[1, 1, 1, 576]].
(nn-Meter) Input shape of mul_15 op is [[1, 7, 7, 576], [1, 7, 7, 576]].
(nn-Meter) Output shape of mul_15 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer11.2.hswish/add op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer11.2.hswish/add op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer11.2.hswish/Relu6.
(nn-Meter) Input shape of layer11.2.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.2.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer11.2.hswish/mul op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer11.2.hswish/mul op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer11.2.hswish/mul_1 op is [[1, 7, 7, 576], [1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.2.hswish/mul_1 op is [[1, 7, 7, 576]].
(nn-Meter) Find node layer11.3.conv/Conv2D with its weight op layer11.3.conv/weight.
(nn-Meter) Get input shape of layer11.3.conv/Conv2D from layer11.2.hswish/mul_1, input shape:[1, 7, 7, 576].
(nn-Meter) Get weight shape of layer11.3.conv/Conv2D from ['layer11.3.conv/weight'], input shape:[1, 1, 576, 96].
(nn-Meter) Op:layer11.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 576], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer11.3.conv/Conv2D op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer11.3.conv/Conv2D op is [[1, 7, 7, 96]].
(nn-Meter) Propagate through op layer11.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer11.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 96]].
(nn-Meter) Output shape of layer11.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 96]].
(nn-Meter) Input shape of Add_12 op is [[1, 7, 7, 96], [1, 7, 7, 96]].
(nn-Meter) Output shape of Add_12 op is [[1, 7, 7, 96]].
(nn-Meter) Find node layer12.1.conv/Conv2D with its weight op layer12.1.conv/weight.
(nn-Meter) Get input shape of layer12.1.conv/Conv2D from Add_12, input shape:[1, 7, 7, 96].
(nn-Meter) Get weight shape of layer12.1.conv/Conv2D from ['layer12.1.conv/weight'], input shape:[1, 1, 96, 576].
(nn-Meter) Op:layer12.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 96], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer12.1.conv/Conv2D op is [[1, 7, 7, 96]].
(nn-Meter) Output shape of layer12.1.conv/Conv2D op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer12.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer12.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer12.1.hswish/add op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer12.1.hswish/add op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer12.1.hswish/Relu6.
(nn-Meter) Input shape of layer12.1.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.1.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer12.1.hswish/mul op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer12.1.hswish/mul op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer12.1.hswish/mul_1 op is [[1, 7, 7, 576], [1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.1.hswish/mul_1 op is [[1, 7, 7, 576]].
(nn-Meter) Find node layer12.2.depconv/depthwise with its weight op layer12.2.depconv/weight.
(nn-Meter) Get input shape of layer12.2.depconv/depthwise from layer12.1.hswish/mul_1, input shape:[1, 7, 7, 576].
(nn-Meter) Get weight shape of layer12.2.depconv/depthwise from ['layer12.2.depconv/weight'], input shape:[5, 5, 576, 1].
(nn-Meter) Op:layer12.2.depconv/depthwise, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 576], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer12.2.depconv/depthwise op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.2.depconv/depthwise op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer12.2.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer12.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.2.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 576]].
(nn-Meter) Get input shape of SE_8/AvgPool from layer12.2.batchnorm/BatchNorm/FusedBatchNormV3, input shape:[1, 7, 7, 576].
(nn-Meter) Op:SE_8/AvgPool, stride:[1, 1, 1, 1], padding:VALID.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 576], kernel size: [1, 7, 7, 1], strides: [1, 1, 1, 1], padding: VALID.
(nn-Meter) Input shape of SE_8/AvgPool op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of SE_8/AvgPool op is [[1, 1, 1, 576]].
(nn-Meter) Find node SE_8/conv2d_16/Conv2D with its weight op SE_8/conv2d_16/kernel.
(nn-Meter) Get input shape of SE_8/conv2d_16/Conv2D from SE_8/AvgPool, input shape:[1, 1, 1, 576].
(nn-Meter) Get weight shape of SE_8/conv2d_16/Conv2D from ['SE_8/conv2d_16/kernel'], input shape:[1, 1, 576, 144].
(nn-Meter) Op:SE_8/conv2d_16/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 576], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_8/conv2d_16/Conv2D op is [[1, 1, 1, 576]].
(nn-Meter) Output shape of SE_8/conv2d_16/Conv2D op is [[1, 1, 1, 144]].
(nn-Meter) Propagate through op SE_8/conv2d_16/BiasAdd.
(nn-Meter) Input shape of SE_8/conv2d_16/BiasAdd op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_8/conv2d_16/BiasAdd op is [[1, 1, 1, 144]].
(nn-Meter) Propagate through op SE_8/Relu.
(nn-Meter) Input shape of SE_8/Relu op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_8/Relu op is [[1, 1, 1, 144]].
(nn-Meter) Find node SE_8/conv2d_17/Conv2D with its weight op SE_8/conv2d_17/kernel.
(nn-Meter) Get input shape of SE_8/conv2d_17/Conv2D from SE_8/Relu, input shape:[1, 1, 1, 144].
(nn-Meter) Get weight shape of SE_8/conv2d_17/Conv2D from ['SE_8/conv2d_17/kernel'], input shape:[1, 1, 144, 576].
(nn-Meter) Op:SE_8/conv2d_17/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 1, 1, 144], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of SE_8/conv2d_17/Conv2D op is [[1, 1, 1, 144]].
(nn-Meter) Output shape of SE_8/conv2d_17/Conv2D op is [[1, 1, 1, 576]].
(nn-Meter) Propagate through op SE_8/conv2d_17/BiasAdd.
(nn-Meter) Input shape of SE_8/conv2d_17/BiasAdd op is [[1, 1, 1, 576]].
(nn-Meter) Output shape of SE_8/conv2d_17/BiasAdd op is [[1, 1, 1, 576]].
(nn-Meter) Input shape of Add_13 op is [[1, 1, 1, 576], []].
(nn-Meter) Output shape of Add_13 op is [[1, 1, 1, 576]].
(nn-Meter) Propagate through op Relu6_8.
(nn-Meter) Input shape of Relu6_8 op is [[1, 1, 1, 576]].
(nn-Meter) Output shape of Relu6_8 op is [[1, 1, 1, 576]].
(nn-Meter) Input shape of mul_16 op is [[1, 1, 1, 576], []].
(nn-Meter) Output shape of mul_16 op is [[1, 1, 1, 576]].
(nn-Meter) Input shape of mul_17 op is [[1, 7, 7, 576], [1, 7, 7, 576]].
(nn-Meter) Output shape of mul_17 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer12.2.hswish/add op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer12.2.hswish/add op is [[1, 7, 7, 576]].
(nn-Meter) Propagate through op layer12.2.hswish/Relu6.
(nn-Meter) Input shape of layer12.2.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.2.hswish/Relu6 op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer12.2.hswish/mul op is [[1, 7, 7, 576], []].
(nn-Meter) Output shape of layer12.2.hswish/mul op is [[1, 7, 7, 576]].
(nn-Meter) Input shape of layer12.2.hswish/mul_1 op is [[1, 7, 7, 576], [1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.2.hswish/mul_1 op is [[1, 7, 7, 576]].
(nn-Meter) Find node layer12.3.conv/Conv2D with its weight op layer12.3.conv/weight.
(nn-Meter) Get input shape of layer12.3.conv/Conv2D from layer12.2.hswish/mul_1, input shape:[1, 7, 7, 576].
(nn-Meter) Get weight shape of layer12.3.conv/Conv2D from ['layer12.3.conv/weight'], input shape:[1, 1, 576, 96].
(nn-Meter) Op:layer12.3.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 576], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of layer12.3.conv/Conv2D op is [[1, 7, 7, 576]].
(nn-Meter) Output shape of layer12.3.conv/Conv2D op is [[1, 7, 7, 96]].
(nn-Meter) Propagate through op layer12.3.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of layer12.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 96]].
(nn-Meter) Output shape of layer12.3.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 96]].
(nn-Meter) Input shape of Add_14 op is [[1, 7, 7, 96], [1, 7, 7, 96]].
(nn-Meter) Output shape of Add_14 op is [[1, 7, 7, 96]].
(nn-Meter) Find node conv13.1.conv/Conv2D with its weight op conv13.1.conv/weight.
(nn-Meter) Get input shape of conv13.1.conv/Conv2D from Add_14, input shape:[1, 7, 7, 96].
(nn-Meter) Get weight shape of conv13.1.conv/Conv2D from ['conv13.1.conv/weight'], input shape:[1, 1, 96, 1024].
(nn-Meter) Op:conv13.1.conv/Conv2D, stride:[1, 1, 1, 1], dilation:[1, 1, 1, 1], padding:SAME.
(nn-Meter) Calculating padding shape, input shape: [1, 7, 7, 96], kernel size: [1, 1], strides: [1, 1, 1, 1], padding: SAME.
(nn-Meter) Input shape of conv13.1.conv/Conv2D op is [[1, 7, 7, 96]].
(nn-Meter) Output shape of conv13.1.conv/Conv2D op is [[1, 7, 7, 1024]].
(nn-Meter) Propagate through op conv11.1.batchnorm/BatchNorm/FusedBatchNormV3.
(nn-Meter) Input shape of conv11.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 1024]].
(nn-Meter) Output shape of conv11.1.batchnorm/BatchNorm/FusedBatchNormV3 op is [[1, 7, 7, 1024]].
(nn-Meter) Input shape of conv13.1.hswish/add op is [[1, 7, 7, 1024], []].
(nn-Meter) Output shape of conv13.1.hswish/add op is [[1, 7, 7, 1024]].
(nn-Meter) Propagate through op conv13.1.hswish/Relu6.
(nn-Meter) Input shape of conv13.1.hswish/Relu6 op is [[1, 7, 7, 1024]].
(nn-Meter) Output shape of conv13.1.hswish/Relu6 op is [[1, 7, 7, 1024]].
(nn-Meter) Input shape of conv13.1.hswish/mul op is [[1, 7, 7, 1024], []].
(nn-Meter) Output shape of conv13.1.hswish/mul op is [[1, 7, 7, 1024]].
(nn-Meter) Input shape of conv13.1.hswish/mul_1 op is [[1, 7, 7, 1024], [1, 7, 7, 1024]].
(nn-Meter) Output shape of conv13.1.hswish/mul_1 op is [[1, 7, 7, 1024]].
(nn-Meter) Get input shape of Mean from conv13.1.hswish/mul_1, input shape:[1, 7, 7, 1024].
(nn-Meter) Get Reduction Indices [1, 2].
(nn-Meter) Input shape of Mean op is [[1, 1024]].
(nn-Meter) Output shape of Mean op is [[1, 1024]].
(nn-Meter) Shape attr find in Reshape op, propagate with normal.
(nn-Meter) Input shape of Reshape op is [[1, 1024]].
(nn-Meter) Output shape of Reshape op is [[-1, 1024]].
(nn-Meter) Find node fc15.fc/MatMul with its weight op fc15.fc/weight.
(nn-Meter) Get weight shape of fc15.fc/MatMul from ['fc15.fc/weight'], input shape:[1024, 1000].
(nn-Meter) Get input shape of fc15.fc/MatMul from Reshape, input shape:[-1, 1024].
(nn-Meter) Input shape of fc15.fc/MatMul op is [[-1, 1024]].
(nn-Meter) Output shape of fc15.fc/MatMul op is [[-1, 1000]].
(nn-Meter) Predict latency: 12.558942703134997 ms
(nn-Meter) [RESULT] predict latency for mobilenetv3small_0.pb: 12.558942703134997 ms
(py37) yongqiang@yongqiang:~$

References

[1] Yongqiang Cheng, https://yongqiang.blog.csdn.net/
[2] microsoft / nn-Meter, https://github.com/microsoft/nn-Meter
[3] nn-meter 2.0, https://pypi.org/project/nn-meter/
[4] nn-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices, https://air.tsinghua.edu.cn/pdf/nn-Meter-Towards-Accurate-Latency-Prediction-of-Deep-Learning-Model-Inference-on-Diverse-Edge-Devices.pdf

Yongqiang Cheng
关注
  • 1
    点赞
  • 踩
  • 1
    收藏
    觉得还不错? 一键收藏
  • 打赏
    打赏
  • 0
    评论
专栏目录
MATLAB算法实战应用案例精讲-【人工智能】语义分割（补充篇）(附matlab代码实现)
qq_36130719的博客
01-01 1742
语义分割是从粗推理到精推理的自然步骤：原点可以定位在分类，分类包括对整个输入进行预测。下一步是本地化/检测，它不仅提供类，还提供关于这些类的空间位置的附加信息。最后，语义分割通过对每个像素进行密集的预测、推断标签来实现细粒度的推理，从而使每个像素都被标记为其封闭对象矿石区域的类别。更具体地说，语义图像分割的目标是给图像中的每个像素都贴上对应的一类所代表的内容的标签。因为我们要对图像中的每个像素进行预测，所以这个任务通常被称为密集预测。
LSTM实现时间序列预测(PyTorch版)
热门推荐
CSDN 精品推荐
01-09 1万+
为了训练数据，首先定义LSTM模型，然后再定义对应的损失函数，由于我们这里是风速预测，显然是个回归问题，所以采用回归问题常用的MESLoss()，如果可以的话，可以自定义损失函数，针对自己的项目需求定义对应的损失函数。对于优化器来讲，使用的也是目前常用的Adam优化器，对于新手来讲也可以多多尝试其它的优化器，比如SGDRMSprop等，对于优化器的选择，可以参考这篇文章Pytorch 30种优化器总结。
Python库 | nn_meter-1.0-py3-none-any.whl
04-24
资源分类：Python库 所属语言：Python 使用前提：需要解压 资源全名：nn_meter-1.0-py3-none-any.whl 资源来源：官方 安装方法：https://lanzao.blog.csdn.net/article/details/101784059
实时性能评估框架nn-Meter解读
csdndevpress0035的博客
07-24 170
本文针对 nn-Meter 框架的关键技术点和可扩展注册接口进行了梳理分析，受限于笔者知识水平，文中可能会存在某些理解身上的偏差，欢迎各位大佬进行交流，共同进步。...
安装和使用 nn-Meter
既然选择了远方 便只顾风雨兼程 - 永强
11-12 409
安装和使用 nn-Meter
Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices
既然选择了远方 便只顾风雨兼程 - 永强
11-21 512
nn-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices nn-Meter：精准预测深度学习模型在边缘设备上的推理延迟
"PCA-DNN模型检测网络异常行为
埃及信息学杂志23（2022）173全文使用PCA-DNN模型检测异常网络行为的Mohammad Al-Fawaa约旦耶尔穆克大学信息技术和计算机科学学院b约旦苏马亚公主技术大学侯赛因国王计算机科学学院c沙特阿拉伯Jouf大学计算机和信息...
【精读文献】1 用于改进脑电图癫痫分析的自监督图神经网络
double_double_的博客
04-12 699
从脑电图中自动检测和分类癫痫可以极大地改善癫痫的诊断和治疗。然而，在先前的自动癫痫检测和分类研究中，有几个建模挑战仍未得到解决：（1）表示脑电图中的非欧几里得数据结构，（2）准确分类罕见的癫痫类型，以及（3）缺乏定量可解释性方法来衡量模型定位癫痫的能力。在这项研究中，我们通过以下方式来应对这些挑战：（1）使用图神经网络（GNN）表示脑电图中的时空依赖性，并提出两种捕捉电极几何形状或动态大脑连接的脑电图图结构；
机器学习数据获取与处理
weixin_73136678的博客
12-15 598
1.5.1 图像处理完整流程图片数据获取图片数据清洗----初步了解数据，筛选掉不合适的图片图片数据标注图片数据预处理data preprocessing。----标准化 standardlization一 中心化 = 去均值 mean normallization一 将各个维度中心化到0一 目的是加快收敛速度,在某些激活函数上表现更好一 归一化 = 除以标准差一 将各个维度的方差标准化处于[-1,1]之间。
顶尖电子称条码秤LS系列称固件，最新固件，固件
08-28
顶尖电子称条码秤LS系列称固件，最新固件，固件
爬取淘宝热销(热门)咖啡商品信息数据集
08-28
本资源旨在收集淘宝平台上热销咖啡商品的相关信息，包括但不限于：店铺所在省份、城市位置、商品名称、销售价格、销量数据、单价（以人民币计）、付款人数、是否提供包邮服务、是否来自天猫平台以及满减优惠的详细信息。这些信息均来源于淘宝平台的公开透明数据，可供学习参考之用。请注意，本资源仅用于学术研究和个人学习目的，不得用于商业用途或其他非法活动。
爬取的淘宝新品推荐男装商品信息公开透明数据集
08-28
本资源专注于收集淘宝平台最新推荐的男装商品信息，内容涵盖商品的店铺所在省份、城市位置、商品的名称、销售价格、累积销量、单价（以人民币计价）、付款的顾客人数、是否提供包邮服务、是否为天猫平台的商品，以及相关的满减优惠情况。这些详细的数据点均来源于淘宝平台的公开透明信息，经过精确抓取和整理，旨在为分析电商平台上的新品推荐策略和消费者购买行为提供实用数据。 这些数据严格遵循淘宝平台的公开政策和隐私保护原则获取，确保了信息的合法性与合规性。然而，本资源仅作为学习参考之用，意在帮助研究人员、市场分析师或学生等理解电商领域的商品推荐机制、销售动态及市场趋势。 任何将此数据用于商业目的或其他未授权的活动都是不恰当的，甚至可能触犯相关法律条款。 在使用这些数据进行学术研究或个人学习时，用户应自觉遵守相关法律法规，尊重数据来源和版权，正确引用数据源，并不得用于任何形式的商业盈利。
STM32F4的环境配置，可直接运行STM32F4系列的芯片
08-28
STM32F4的环境配置，可直接运行STM32F4系列的芯片
技术资料PCB资料模拟电子技术资料03j.zip
08-28
技术资料PCB资料模拟电子技术资料03j.zip
全国大学生电子设计大赛项目合集射频无线资料雷达接收机灵敏度自动测试方案设计
08-28
全国大学生电子设计大赛项目合集射频无线资料雷达接收机灵敏度自动测试方案设计提取方式是百度网盘分享地址
da_1724814826825 (1)
08-28
da_1724814826825 (1)
单片机实现的超声波测距与显示模块.md
最新发布
08-28
如何利用单片机实现超声波测距与显示模块，包括从硬件设计到软件编程的全方位内容。首先，介绍了超声波测距的基本原理，使用超声波模块（如HC-SR04）通过发射和接收超声波信号来测量距离，并通过公式计算距离。接着，阐述了单片机（如STC89C52、ATmega328）如何与超声波模块连接，以及电路设计细节。软件部分涵盖了如何编写单片机程序以触发超声波发射、测量回波时间，并计算距离。代码示例展示了如何使用计时器测量回波时间，并通过LCD显示距离。最后，文章还包括了调试与优化建议，如硬件连接检查、软件调试方法及性能优化策略，以确保系统的准确性和稳定性。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

Yongqiang Cheng

梦想不是浮躁，而是沉淀和积累。

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值