notebooks and models

DockerLSTM/Dockerfile

@@ -0,0 +1,16 @@
+FROM continuumio/miniconda
+
+RUN pip install tensorflow
+RUN pip install keras==1.2.2
+RUN pip install h5py
+RUN pip install flask
+RUN pip install cherrypy
+
+EXPOSE 8002
+
+RUN mkdir /webapp
+ADD . /webapp
+
+WORKDIR /webapp
+
+CMD python /webapp/server.py

DockerLSTM/app.py

@@ -0,0 +1,68 @@
+import pickle
+import os
+import json
+import random
+import numpy as np
+import tensorflow as tf
+from flask import Flask, request
+from keras.preprocessing.text import Tokenizer
+from keras.preprocessing.sequence import pad_sequences
+from keras.models import load_model
+
+app = Flask(__name__)
+
+features = ['cool', 'funny', 'stars', 'useful']
+
+tokenizers = {}
+models = {}
+graphs = {}
+
+for feature in features:
+	with open('/webapp/' + feature + '/' + feature + '.tokenizer', 'rb') as f:
+		tokenizers.update({feature: pickle.load(f)})
+
+	models.update({feature: load_model('/webapp/' + feature + '/' + feature + '.model')})
+	# https://github.com/fchollet/keras/issues/2397
+	graphs.update({feature: tf.get_default_graph()})
+
+def make_pred(feature, sentence):
+	sequence = tokenizers[feature].texts_to_sequences([sentence])
+	padded_seq = pad_sequences(sequence, maxlen=400)
+	with graphs[feature].as_default():
+		predicted_proba = models[feature].predict_proba(padded_seq)
+	return predicted_proba
+
+@app.route("/cool", methods=["POST"])
+def cool():
+	sentence = request.form.get('review').encode('utf-8')
+	predicted_proba = make_pred('cool', sentence)
+	prediction = (predicted_proba > 0.5).astype(int)[0][0]
+	response = {'cool': prediction}
+	return json.dumps(response)
+
+@app.route("/funny", methods=["POST"])
+def funny():
+	sentence = request.form.get('review').encode('utf-8')
+	predicted_proba = make_pred('funny', sentence)
+	prediction = (predicted_proba > 0.5).astype(int)[0][0]
+	response = {'funny': prediction}
+	return json.dumps(response)
+
+@app.route("/stars", methods=["POST"])
+def stars():
+	sentence = request.form.get('review').encode('utf-8')
+	predicted_proba = make_pred('stars', sentence)
+	prediction = np.argmax(predicted_proba, axis=1)[0]
+	response = {'stars': prediction}
+	return json.dumps(response)
+
+@app.route("/useful", methods=["POST"])
+def useful():
+	sentence = request.form.get('review').encode('utf-8')
+	predicted_proba = make_pred('useful', sentence)
+	prediction = (predicted_proba > 0.5).astype(int)[0][0]
+	response = {'useful': prediction}
+	return json.dumps(response)
+
+if __name__ == "__main__":
+    app.run(port=8002)