tensorflow 学习笔记 1

安装完tensorflow后，执行

source activate tensorflow

测试mnist手写识别example：

:~/mnist$ python fully_connected_feed.py --input_data_dir .
Extracting ./train-images-idx3-ubyte.gz
Extracting ./train-labels-idx1-ubyte.gz
Extracting ./t10k-images-idx3-ubyte.gz
Extracting ./t10k-labels-idx1-ubyte.gz
2017-12-04 21:26:45.617122: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
Step 0: loss = 2.33 (0.179 sec)
Step 100: loss = 2.17 (0.004 sec)
Step 200: loss = 1.94 (0.004 sec)
Step 300: loss = 1.65 (0.004 sec)
Step 400: loss = 1.29 (0.004 sec)
Step 500: loss = 0.97 (0.004 sec)
Step 600: loss = 0.86 (0.004 sec)
Step 700: loss = 0.61 (0.004 sec)
Step 800: loss = 0.60 (0.004 sec)
Step 900: loss = 0.55 (0.009 sec)
Training Data Eval:
  Num examples: 55000  Num correct: 47555  Precision @ 1: 0.8646
Validation Data Eval:
  Num examples: 5000  Num correct: 4358  Precision @ 1: 0.8716
Test Data Eval:
  Num examples: 10000  Num correct: 8676  Precision @ 1: 0.8676
Step 1000: loss = 0.54 (0.013 sec)
Step 1100: loss = 0.55 (0.106 sec)
Step 1200: loss = 0.44 (0.004 sec)
Step 1300: loss = 0.40 (0.004 sec)
Step 1400: loss = 0.52 (0.003 sec)
Step 1500: loss = 0.36 (0.006 sec)
Step 1600: loss = 0.42 (0.004 sec)
Step 1700: loss = 0.51 (0.004 sec)
Step 1800: loss = 0.56 (0.003 sec)
Step 1900: loss = 0.35 (0.004 sec)
Training Data Eval:
  Num examples: 55000  Num correct: 49276  Precision @ 1: 0.8959
Validation Data Eval:
  Num examples: 5000  Num correct: 4504  Precision @ 1: 0.9008
Test Data Eval:
  Num examples: 10000  Num correct: 8989  Precision @ 1: 0.8989

启动tensorflow dashboard

:~/mnist$ tensorboard --logdir=log
TensorBoard 0.4.0rc2 at http://ardell-virtual-machine:6006 (Press CTRL+C to quit)

浏览器里访问port 6006，得到graph截图：

完全不会看这张图，希望下次会好点