TensorFlow MNIST

This repository is going to use TensorFlow to train the MNIST network.

File Structure

mnist/
    |--- out/                       # For the ouput files
    |--- src/                       # Source code are in this directory
        |--- mnist_fully.py         # 2-Hidden Layer Fully Connected NN with TensorFlow
        |--- mnist_fully_model.py   # Add save and restore in mnist_fully.py
    |--- README.md      
    |--- LICENSE
    |--- .gitignore

Installation

The following instructions are for installing on Ubuntu Linux 16.04

TensorFlow

TensorFlow is tested and supported on the following 64-bit systems:

Ubuntu 16.04 or later

Windows 7 or later

macOS 10.12.6 (Sierra) or later (no GPU support)

Raspbian 9.0 or later

Install TensorFlow

For Python 2.7, you can follow the instructions here.

Prerequisite (for Python 3 (Python 3.4, 3.5, 3.6))

# Check if your Python environment is already configured
$ python3 --version
$ pip3 --version
$ virtualenv --version
# If the above packages are already installed, skip to the next step
$ sudo apt-get update
$ sudo apt-get install python3-dev python3-pip
# Install for system-wide
$ sudo pip3 install -U virtualenv

Create a virtual environment (recommended)

# Create a new vitrual environment by choosing a Python interpreter and making a ./env directory to hold it
$ virtualenv --system-site-packages -p python3 ./venv
# Activate the virtual environement using a shell-specific command (e.g., sg, bash, etc.)
$ source ./venv/bin/activate
# When virtualenv is active, your shell prompt is prefixed with (venv).
(venv) $
# Install packages within a virtual environment without affecting the host system setup. Start by upgrading pip:
(venv) $ pip install --upgrade pip
# Show packages installed within the virtual environment
(venv) $ pip list
# To exit virtualenv later
(venv) $ deactivate

Install TensorFlow with Python's pip package manager

# Current release for GPU-only (Python 2.7)
(venv) $ pip install --upgrade tensorflow
# GPU package for CUDA-enabled GPU cards (Python 2.7)
(venv) $ pip install --upgrade tensorflow-gpu

Run the example program hello.py

# Make sure your current directory is src/
(venv) $ python hello.py
b'Hello, TensorFlow!'

Execution

./src/mnist_fully.py

$ python mnist_fully.py
2018-12-11 18:57:18.102112: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
Step 1, Minibatch Loss= 8470.1895, Training Accuracy= 0.328
Step 100, Minibatch Loss= 250.3591, Training Accuracy= 0.875
Step 200, Minibatch Loss= 173.7493, Training Accuracy= 0.828
Step 300, Minibatch Loss= 155.4185, Training Accuracy= 0.820
Step 400, Minibatch Loss= 81.8946, Training Accuracy= 0.859
Step 500, Minibatch Loss= 18.5791, Training Accuracy= 0.938
Optimization Finished!
Testing Accuracy: 0.8563

./src/mnist_fully_model.py

$ python mnist_fully_model.py
2018-12-11 19:18:36.799674: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
Step 1, Minibatch Loss= 9184.1309, Training Accuracy= 0.391
Step 100, Minibatch Loss= 365.5018, Training Accuracy= 0.852
Step 200, Minibatch Loss= 153.3021, Training Accuracy= 0.820
Step 300, Minibatch Loss= 24.0616, Training Accuracy= 0.898
Step 400, Minibatch Loss= 56.0503, Training Accuracy= 0.875
Step 500, Minibatch Loss= 74.1631, Training Accuracy= 0.844
Optimization Finished!
Model saved in file: /tmp/model.ckpt
Model restored from file: /tmp/model.ckpt
Testing Accuracy: 0.8559

./src/mnist_fully_input.py

$ python mnist_fully_input.py
2018-12-11 19:25:24.784290: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
Model restored from file: /tmp/model.ckpt
Answer: [7 2 1 ..., 4 5 6]

Contributor

David Lu

License

Apache License 2.0