如何查看tf SavedModel的輸入/輸出等信息？

參考鏈接：https://juejin.im/post/6844903693184172040

查看模型的Signature簽名

Tensorflow提供了一個工具

• 如果你下載了Tensorflow的源碼，可以找到這樣一個文件，./tensorflow/python/tools/saved_model_cli.py
• 如果你安裝了tensorflow，也可以用下邊的命令查看tensorflow源碼位置和版本：

import tensorflow as tf
print tf.__path__
print tf.__version__

你可以加上-h參數查看saved_model_cli.py腳本的幫助信息：

usage: saved_model_cli.py [-h] [-v] {show,run,scan} ...

saved_model_cli: Command-line interface for SavedModel

optional arguments:
  -h, --help       show this help message and exit
  -v, --version    show program's version number and exit

commands:
  valid commands

  {show,run,scan}  additional help

如果你安裝

指定SavedModel模所在的位置，我們就可以顯示SavedModel的模型信息：

python path/to/tensorflow/python/tools/saved_model_cli.py show --dir ./model/ --all

顯示類似結果

MetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:

signature_def['predict']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['myInput'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 784)
        name: myInput:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['myOutput'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 10)
        name: Softmax:0
  Method name is: tensorflow/serving/predict

查看模型的計算圖

了解tensflow的人可能知道TensorBoard是一個非常強大的工具，能夠顯示很多模型信息，其中包括計算圖。問題是，TensorBoard需要模型訓練時的log，如果這個SavedModel模型是別人訓練好的呢？辦法也不是沒有，我們可以寫一段代碼，加載這個模型，然后輸出summary info，代碼如下：

import tensorflow as tf
import sys
from tensorflow.python.platform import gfile

from tensorflow.core.protobuf import saved_model_pb2
from tensorflow.python.util import compat

with tf.Session() as sess:
  model_filename ='./model/saved_model.pb'
  with gfile.FastGFile(model_filename, 'rb') as f:

    data = compat.as_bytes(f.read())
    sm = saved_model_pb2.SavedModel()
    sm.ParseFromString(data)

    if 1 != len(sm.meta_graphs):
      print('More than one graph found. Not sure which to write')
      sys.exit(1)

    g_in = tf.import_graph_def(sm.meta_graphs[0].graph_def)
LOGDIR='./logdir'
train_writer = tf.summary.FileWriter(LOGDIR)
train_writer.add_graph(sess.graph)
train_writer.flush()
train_writer.close()

代碼中，將匯總信息輸出到logdir，接着啟動TensorBoard，加上上面的logdir:

tensorboard --logdir ./logdir

在瀏覽器中輸入地址： http://127.0.0.1:6006/ ，就可以看到如下的計算圖：