Initial version of the workshop.

.gitignore

@@ -8,6 +8,8 @@ __pycache__/
 
 # Distribution / packaging
 .Python
+
+env/
 build/
 develop-eggs/
 dist/
@@ -80,15 +82,13 @@ celerybeat-schedule
 
 # SageMath parsed files
 *.sage.py
-
-# Environments
+# dotenv
 .env
+
+# virtualenv
 .venv
-env/
 venv/
 ENV/
-env.bak/
-venv.bak/
 
 # Spyder project settings
 .spyderproject
@@ -102,3 +102,13 @@ venv.bak/
 
 # mypy
 .mypy_cache/
+
+**/MNIST_data
+**/logs
+
+**/checkpoint
+**/mnist_intermediate*
+**/.DS_Store
+**/17flowers
+**/model_output
+*.h5

Dockerfile

@@ -0,0 +1,15 @@
+FROM jupyter/scipy-notebook
+
+MAINTAINER Wilder Rodrigues ([email protected])
+
+USER $NB_USER
+
+# Install TensorFlow
+RUN conda install -c conda-forge tensorflow -y && \
+    conda install -c conda-forge numpy keras nltk gensim -y
+
+# Install Reinforcement Learning packages:
+RUN pip install gym==0.9.4
+
+# Install Keras Kontrib
+RUN pip install git+https://www.github.com/keras-team/keras-contrib.git

Dockerfile-gpu

@@ -0,0 +1,17 @@
+FROM ekholabs/nvidia-cuda
+FROM jupyter/scipy-notebook
+
+MAINTAINER Wilder Rodrigues <[email protected]>
+
+USER $NB_USER
+
+RUN conda install -c conda-forge tensorflow-gpu -y && \
+    conda install -c conda-forge numpy keras nltk -y
+
+# Install Reinforcement Learning packages:
+RUN pip install gym==0.9.4
+
+# Install Keras Kontrib
+RUN pip install git+https://www.github.com/keras-team/keras-contrib.git
+
+EXPOSE 8888

README.md

@@ -0,0 +1,37 @@
+# Workshop
+
+The examples cover a straightforward start, from shallow to intermediate, deep and CNN networks. It also shows how the trained model can be tested and evaluated.
+
+## Build Docker Image
+
+After cloning this repository, please execute the command below to build the Docker image.
+
+```
+docker build -t ekholabs/deeplearning-stack .
+```
+
+## Run Docker Container
+
+Once you have built the image, please execute the command below to run the container.
+
+```
+docker run -v [path_to_project]/workshop:/home/jovyan/work --rm -p 8888:8888 ekholabs/deeplearning-stack
+```
+
+* Remark: 'jovyan' is the default Docker user.
+
+## Jupyter Notebooks
+
+After starting the Docker container, copy the Jupyter notebook URL and start working.
+
+* Remark: if you face problems concerning lack of resources, please increase your Docker Engine memory. I tested the notebooks in a MacBook Pro with 16GB of RAM. I dedicated 5GB to my Docker Engine.
+
+## TensorBoard
+
+If you want to visualise the loss and accuracy metrics, just execute TensorBoard pointing to your logs directory:
+
+```
+tensorboard --logdir [path_to_project]/DLinK/notebooks/logs
+````
+
+* Remark: the 'logs' directory is not part of the repository. It has to be created under the 'notebooks' directory. All the Jupyter notebook are already configured to use 'notebooks/logs' for the TensorBoard files.

notebooks/keras/alexnet-in-keras.ipynb

@@ -0,0 +1,197 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## AlexNet"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Oxford Flowers classification"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Set the seed"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "np.random.seed(42)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Load dependencies"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import keras\n",
+    "from keras.models import Sequential\n",
+    "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D\n",
+    "from keras.layers.normalization import BatchNormalization\n",
+    "from keras.callbacks import TensorBoard"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Load and Process the data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import tflearn.datasets.oxflower17 as oxflower17\n",
+    "X, Y = oxflower17.load_data(one_hot = True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Design Neural Network Architecture"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model = Sequential()\n",
+    "\n",
+    "# Kernel size is 3x3\n",
+    "# Default strides = (1, 1)\n",
+    "model.add(Conv2D(64, 3, activation = \"relu\", input_shape = (224, 224, 3)))\n",
+    "model.add(Conv2D(64, 3, activation = \"relu\"))\n",
+    "model.add(MaxPooling2D((2, 2)))\n",
+    "model.add(BatchNormalization())\n",
+    "\n",
+    "model.add(Conv2D(128, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(128, 3, activation = \"relu\"))\n",
+    "model.add(MaxPooling2D((2, 2)))\n",
+    "model.add(BatchNormalization())\n",
+    "          \n",
+    "model.add(Conv2D(256, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(256, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(256, 3, activation = \"relu\"))\n",
+    "model.add(MaxPooling2D((2, 2)))\n",
+    "model.add(BatchNormalization())\n",
+    "\n",
+    "model.add(Conv2D(512, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(512, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(512, 3, activation = \"relu\"))\n",
+    "model.add(MaxPooling2D((2, 2)))\n",
+    "model.add(BatchNormalization())\n",
+    "\n",
+    "model.add(Conv2D(512, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(512, 3, activation = \"relu\"))\n",
+    "model.add(Conv2D(512, 3, activation = \"relu\"))\n",
+    "model.add(MaxPooling2D((2, 2)))\n",
+    "model.add(BatchNormalization())\n",
+    "\n",
+    "model.add(Flatten())\n",
+    "model.add(Dense(4096, activation = \"tanh\"))\n",
+    "model.add(Dropout(0.5))\n",
+    "model.add(Dense(4096, activation = \"tanh\"))\n",
+    "model.add(Dropout(0.5))\n",
+    "\n",
+    "model.add(Dense(17, activation = \"softmax\"))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.summary()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.compile(loss = \"categorical_crossentropy\", optimizer = \"adam\", metrics = [\"accuracy\"])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### TensorBoard"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "tensorboard = TensorBoard(\"logs/alexnet\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.fit(X, Y, batch_size = 64, epochs = 1, verbose = 1, validation_split = 0.1, shuffle = True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}