Day #121 - Tensorflow Notes

Estimators

• Provide Boiler Plate code
• tf.estimator - High level API for production ready models
• Python API treats tensorflow as numeric processing library

Estimator API

• Quick Model
• Checkpointing
• Distributed Training
• Train / Eval / Monitor
• Out of memory datasets
• Hyper parameter tuning

Base class

• tf.estimator.Estimator
• Linear (LinearRegressor)
• Dense Neural Networks (DNNRegressor, DNNLinearCombinedRegressor)
• LinearClassifier, DNNClassifier, DNNLinearCombinedClassifier

	import tensorflow as tf
	categorical_column = tf.feature_column.categorical_column_with_vocabulary_list(key="type", vocabulary_list=["house","apt"])
	featcols = [
	tf.feature_column.numeric_column("sq_footage"),
	tf.feature_column.indicator_column(categorical_column)
	]
	initializer=tf.zeros_initializer()
	#featcols = [
	#tf.feature_column.numeric_column("sq_footage"),
	#tf.feature_column.categorical_column_with_vocabulary_list("type",["house","apt"])
	#]
	#model = tf.estimator.LinearRegressor(featcols)
	#instantiate linear regressor based on feilds
	#3 and 2 Neurons
	model = tf.estimator.DNNRegressor(feature_columns=featcols,hidden_units=[3,2])
	#model = tf.estimator.DNNRegressor(feature_columns=featcols, hidden_units=[3,2],dropout=0.2,optimizer="Adam")
	#LinearRegressor - computes weighted sum of all input elements
	#Many more options
	#tf.feature_column.bucketized_column
	#tf.feature_column.embedding_column
	#tf.feature_column.crossed_column
	#tf.feature_column.categorical_column_with_hash_bucket
	# 3 houses, 3 apartments
	def train_input_fn():
	features = {"sq_footage":[1000,2000,3000,1000,2000,3000],"type":["house","house","house","apt","apt","apt"]}
	labels = [500,1000,1500,700,1300,1900]
	return features, labels
	#trains on this batch
	#train for 100 steps
	model.train(train_input_fn,steps=100)
	def predict_input_fn():
	features = {"sq_footage":[1500,1800], "type":["house","apt"]}
	return features
	predictions = model.predict(predict_input_fn)
	#to print values
	print(next(predictions))

view raw tensor2.py hosted with ❤ by GitHub

Happy Learning!!!