相关疑难解决方法(0)

主要思想是将TFRecords转换为numpy数组.假设TFRecord存储图像.特别:

1. 读取TFRecord文件并将每个图像转换为numpy数组.
2. 将图像写入1.jpg,2.jpg等
3. 同时,将文件名和标签写入文本文件,如下所示:

   1.jpg 2
   2.jpg 4
   3.jpg 5
   

   Run Code Online (Sandbox Code Playgroud)

我目前使用以下代码:

import tensorflow as tf
import os

def read_and_decode(filename_queue):
  reader = tf.TFRecordReader()
  _, serialized_example = reader.read(filename_queue)
  features = tf.parse_single_example(
      serialized_example,
      # Defaults are not specified since both keys are required.
      features={
          'image_raw': tf.FixedLenFeature([], tf.string),
          'label': tf.FixedLenFeature([], tf.int64),
          'height': tf.FixedLenFeature([], tf.int64),
          'width': tf.FixedLenFeature([], tf.int64),
          'depth': tf.FixedLenFeature([], tf.int64)
      })
  image = tf.decode_raw(features['image_raw'], tf.uint8)
  label = tf.cast(features['label'], tf.int32)
  height = tf.cast(features['height'], tf.int32)
  width = tf.cast(features['width'], tf.int32)
  depth = tf.cast(features['depth'], tf.int32)
  return image, label, …

我正在使用Keras和Tensorflow实施强化学习模型。我必须在单个输入上频繁调用model.predict（）。

在对简单的预训练模型进行推理测试时，我注意到使用Keras的model.predict比仅对存储的权重使用Numpy慢。为什么这么慢，我如何加速呢？对于复杂模型，使用纯Numpy不可行。

import timeit
import numpy as np
from tensorflow.python.keras.models import Sequential
from tensorflow.python.keras.layers import Dense

w = np.array([[-1., 1., 0., 0.], [0., 0., -1., 1.]]).T
b = np.array([ 15., -15., -21., 21.])

model = Sequential()
model.add(Dense(4, input_dim=2, activation='linear'))
model.layers[0].set_weights([w.T, b])
model.compile(loss='mse', optimizer='adam')

state = np.array([-23.5, 17.8])

def predict_very_slow():
    return model.predict(state[np.newaxis])[0]

def predict_slow():
    ws = model.layers[0].get_weights()
    return np.matmul(ws[0].T, state) + ws[1]

def predict_fast():
    return np.matmul(w, state) + b

print(
    timeit.timeit(predict_very_slow, number=10000),
    timeit.timeit(predict_slow, number=10000),
    timeit.timeit(predict_fast, number=10000)
)
# 5.168972805004538 1.6963867129435828 0.021918574168087623 …

我正在尝试将我们的输入管道移动到 tensorflow 数据集 api。为此，我们已将图像和标签转换为 tfrecords。然后我们通过dataset api读取tfrecords，比较初始数据和读取的数据是否相同。到现在为止还挺好。下面是将 tfrecords 读入数据集的代码

def _parse_function2(proto):

    # define your tfrecord again. Remember that you saved your image as a string.

    keys_to_features = {"im_path": tf.FixedLenSequenceFeature([], tf.string, allow_missing=True),
                        "im_shape": tf.FixedLenSequenceFeature([], tf.int64, allow_missing=True),
                        "score_shape": tf.FixedLenSequenceFeature([], tf.int64, allow_missing=True),
                        "geo_shape": tf.FixedLenSequenceFeature([], tf.int64, allow_missing=True),
                        "im_patches": tf.FixedLenSequenceFeature([], tf.string, allow_missing=True),
                        "score_patches": tf.FixedLenSequenceFeature([], tf.string, allow_missing=True),
                        "geo_patches": tf.FixedLenSequenceFeature([], tf.string, allow_missing=True),
                        }

    # Load one example
    parsed_features = tf.parse_single_example(serialized=proto, features=keys_to_features)

    parsed_features['im_patches'] = parsed_features['im_patches'][0]
    parsed_features['score_patches'] = parsed_features['score_patches'][0]
    parsed_features['geo_patches'] = parsed_features['geo_patches'][0]

    parsed_features['im_patches'] = tf.decode_raw(parsed_features['im_patches'], tf.uint8)
    parsed_features['im_patches'] = tf.reshape(parsed_features['im_patches'], parsed_features['im_shape']) …