def __init__(self, data_location=None):
self.arg_obj = ArgParser()
self.data_location = data_location
self.sampler = None
self.feature_extraction_mapper = None
self.feature_extraction_reducer = None
self.mapper_param = None
self.reducer_param = None
self.training_set_size = None
self.test_set_size = None
def __init__(self, data_location=None, model_file=None):
self.arg_obj = ArgParser()
# this is the local location where test dataset is located
self.data_location = data_location
self.model_file = model_file
self.classification_mapper = None
self.classification_reducer = None
self.mapper_param = None
self.reducer_param = None
class SamplingInterface:
def __init__(self):
self.arg_obj = ArgParser()
# method to show help message
def show_help(self):
print ":( not enough params"
print " usage: python sample_interface.py -sampler <sampler_implementation> -positive <positive_source_file> -negative <negative_source_file> -train_size <train_set_size> -test_size <test_set_size>"
exit()
# method that checks if required parameters are there or not
# returns False if the required params are missing
# returns True if all the required params are provided
def check_params(self):
print "sampler checking:",self.arg_obj.args
if not self.arg_obj.args.has_key("sampler") or not self.arg_obj.args.has_key("positive") or not self.arg_obj.args.has_key("negative") or not self.arg_obj.args.has_key("train_size") or not self.arg_obj.args.has_key("test_size"):
self.show_help()
else:
return True
def sample(self):
self.arg_obj.parse(sys.argv)
# check if required parameters are provided
self.check_params()
sampler_implementation = self.arg_obj.args["sampler"]
positive_source_file = self.arg_obj.args["positive"]
negative_source_file = self.arg_obj.args["negative"]
train_set_size = self.arg_obj.args["train_size"]
test_set_size = self.arg_obj.args["test_size"]
# this dynamically loads the concrete implementation of sampler
_dyn_cls_loader = DynamicClassLoader()
_class_placeholder = _dyn_cls_loader.load(sampler_implementation)
print "SamplingInterface:Instantiating:",_class_placeholder
# instantiating the sampler
_instance = _class_placeholder(positive_source_file, negative_source_file)
_instance.generateDatasets(train_set_size, test_set_size)
print "SamplingStub: sampling done."
class ModelingStub:
def __init__(self):
self.arg_obj = ArgParser()
# method that checks if required parameters are there or not
# returns False if the required params are missing
# returns True if all the required params are provided
def check_params(self):
print "modeler checking:",self.arg_obj.args
if not self.arg_obj.args.has_key("feature_set_location"):
return False
else:
return True
# method to show error message
def show_help(self):
print ":( not enough params"
print "usage: python modeler_interface.py -feature_set_location <feature_set_file_location> -model_name <model_name_to_save_as>"
print "==or=="
print "usage: python modeler_interface.py -feature_set_location <feature_set_location_directory>"
exit()
# start the modeling using the feature set
def model(self):
self.arg_obj.parse(sys.argv)
if not self.check_params():
self.show_help()
_feature_set_location = self.arg_obj.args["feature_set_location"]
if os.path.isfile(_feature_set_location):
# if given a file path and not provided the model name to save as
if not self.arg_obj.args.has_key("model_name"):
self.show_help()
_model_name = self.arg_obj.args["model_name"]
_instance = POSContextSequenceModeler(feature_set_location = _feature_set_location)
_instance.train()
_instance.save_model(name=_model_name,location="trained_models")
print "ModelingStub: modeling done for given feature set file."
if os.path.isdir(_feature_set_location):
print "ModelingStub: looking into feature set directory..."
# filter only feature set files with .txt extension
file_list = [fn for fn in os.listdir(_feature_set_location) if fn.endswith(('.txt'))]
for _file in file_list:
_path = _feature_set_location+"/"+_file
_coll = _file.split(".")
_model_name = _coll[0]+".model"
_instance = POSContextSequenceModeler(feature_set_location = _path)
_instance.train()
print "ModelingStub: trained the model.about to save."
_instance.save_model(name=_model_name,location="trained_models")
print "ModelingStub: modeling done for:",_file
print "ModelingStub: modeling done for all files in directory provided."
def __init__(self):
self.arg_obj = ArgParser()
def __init__(self):
self.arg_obj = ArgParser()
class ModelingStub:
def __init__(self):
self.arg_obj = ArgParser()
# method that checks if required parameters are there or not
# returns False if the required params are missing
# returns True if all the required params are provided
def check_params(self):
print "modeler checking:", self.arg_obj.args
if not self.arg_obj.args.has_key("feature_set_location"):
return False
else:
return True
# method to show error message
def show_help(self):
print ":( not enough params"
print "usage: python modeler_interface.py -feature_set_location <feature_set_file_location> -model_name <model_name_to_save_as>"
print "==or=="
print "usage: python modeler_interface.py -feature_set_location <feature_set_location_directory>"
exit()
# start the modeling using the feature set
def model(self):
self.arg_obj.parse(sys.argv)
if not self.check_params():
self.show_help()
_feature_set_location = self.arg_obj.args["feature_set_location"]
if os.path.isfile(_feature_set_location):
# if given a file path and not provided the model name to save as
if not self.arg_obj.args.has_key("model_name"):
self.show_help()
_model_name = self.arg_obj.args["model_name"]
_instance = POSContextSequenceModeler(
feature_set_location=_feature_set_location)
_instance.train()
_instance.save_model(name=_model_name, location="trained_models")
print "ModelingStub: modeling done for given feature set file."
if os.path.isdir(_feature_set_location):
print "ModelingStub: looking into feature set directory..."
# filter only feature set files with .txt extension
file_list = [
fn for fn in os.listdir(_feature_set_location)
if fn.endswith(('.txt'))
]
for _file in file_list:
_path = _feature_set_location + "/" + _file
_coll = _file.split(".")
_model_name = _coll[0] + ".model"
_instance = POSContextSequenceModeler(
feature_set_location=_path)
_instance.train()
print "ModelingStub: trained the model.about to save."
_instance.save_model(name=_model_name,
location="trained_models")
print "ModelingStub: modeling done for:", _file
print "ModelingStub: modeling done for all files in directory provided."
class FeatureExtractionInterface:
def __init__(self, data_location=None):
self.arg_obj = ArgParser()
self.data_location = data_location
self.sampler = None
self.feature_extraction_mapper = None
self.feature_extraction_reducer = None
self.mapper_param = None
self.reducer_param = None
self.training_set_size = None
self.test_set_size = None
# method to show help message
def show_help(self):
print ":( not enough params"
print "usage: python feature_extraction_interface.py -fe_mapper <feature_extraction_mapper> -fe_mapper_params <mapper_params> -fe_reducer <feature_extraction_reducer> -fe_reducer_params <reducer_params> -train_dataset <dataset_location> -train_size <train_set_size> -test_size <test_set_size>"
exit()
# method that checks if required parameters are there or not
# returns False if the required params are missing
# returns True if all the required params are provided
def check_params(self):
# mapper and reducer params might be optional- thus they are not required
if not self.arg_obj.args.has_key("fe_mapper") or not self.arg_obj.args.has_key("fe_reducer") or not self.arg_obj.args.has_key("train_dataset") or not self.arg_obj.args.has_key("train_size") or not self.arg_obj.args.has_key("test_size") :
self.show_help()
exit()
# method that invokes sampling- it assumes that positive instances file is named 'positive_instances' and negative instances file is named 'negative_instances'
def invoke_sampling(self):
self.check_params()
# do sampling
# using default sampler if not provided sampler in param list
if not self.arg_obj.args.has_key("sampler"):
_sampler = "lib.sampler.random_sampler.RandomSampler"
else:
_sampler = self.arg_obj.args["sampler"]
_cmd = "python "+_prefix+"/"+"sampler_interface.py "+self.arg_obj.get_string()
os.system(_cmd)
time.sleep(5)
# method that removes the dataset directory on HDFS
def remove_dataset_dir_on_hdfs(self):
self.check_params()
# remove training directory on HDFS
_cmd = "hadoop fs -rmr /user/hadoop/feature_extraction_input"
os.system(_cmd)
time.sleep(5)
# method that removes the output directory on HDFS
def remove_output_dir_on_hdfs(self):
self.check_params()
# remove output directory on HDFS
_cmd = "hadoop fs -rmr /user/hadoop/feature_extraction_output"
os.system(_cmd)
time.sleep(5)
# method that loads the dataset into HDFS
def load_data_set_on_hdfs(self):
self.check_params()
# load new training data on HDFS
_cmd = "hadoop fs -put Train/train_set_w_tags /user/hadoop/feature_extraction_input/"
os.system(_cmd)
time.sleep(10)
# method that starts the feature extraction job
def start_feature_extraction_job(self):
self.check_params()
print "Launching map-reduce feature extraction task..."
# start feature extraction
_cmd = "hadoop jar /home/hadoop/contrib/streaming/hadoop-streaming-1.0.3.jar -input /user/hadoop/feature_extraction_input -mapper '"+ self.feature_extraction_mapper
# use parameters for mapper job if they are provided
if not self.mapper_param is None:
_cmd = _cmd +" "+ self.mapper_param
_cmd = _cmd + "' -file "+ self.feature_extraction_mapper +" -reducer '"+ self.feature_extraction_reducer
# use parameters for reducer job if they are provided
if not self.reducer_param is None:
_cmd = _cmd +" "+ self.reducer_param
_cmd = _cmd + "' -file "+ self.feature_extraction_reducer +" -file glossextractionengine.mod -output /user/hadoop/feature_extraction_output -jobconf mapred.job.name='GlossExtractionEngine:FeatureExtraction'"
os.system(_cmd)
time.sleep(5)
print "feature extraction task completed."
# method that exports the result of feature extraction from HDFS to local file system
def export_output_from_hdfs(self):
self.check_params()
# create the output directory for featuer extraction job
if not os.path.exists("feature_set_for_modeling"):
os.system("mkdir feature_set_for_modeling")
# remove previous version of the feature set file in the output directory
if os.path.exists("./feature_set_for_modeling/"+str(self.training_set_size)+"_output.txt"):
print "FeatureExtractionInterface: File already exists.. removing it :","./feature_set_for_modeling/"+str(self.training_set_size)+"_output.txt"
os.remove("./feature_set_for_modeling/"+str(self.training_set_size)+"_output.txt")
# get the merged output from HDFS
_cmd = "hadoop fs -getmerge /user/hadoop/feature_extraction_output ./feature_set_for_modeling/"+str(self.training_set_size)+"_output.txt"
os.system(_cmd)
print "Saved output[Feature set for modeling] at : feature_set_for_modeling/"+str(self.training_set_size)+"_output.txt"
# method to perform sequence of operations before launching a map-reduce job for feature extraction
def launch(self):
self.check_params()
if self.arg_obj.args.has_key("sampler"):
# interact with sampling interface for sampling
self.invoke_sampling()
self.remove_dataset_dir_on_hdfs()
self.remove_output_dir_on_hdfs()
self.load_data_set_on_hdfs()
# start the feature extraction job
self.start_feature_extraction_job()
self.export_output_from_hdfs()
class InterfaceWrapper:
def __init__(self):
self.arg_obj = ArgParser()
# method to show help message
def show_help(self):
print ":( not enough params"
print "usage: python run.py -operation <operation_name> <parameters for operation>"
print "******supported operations******"
print "(1) operation name: sampling"," parameters: -sampler <sampler_implementation> -positive <positive_source_file> -negative <negative_source_file> -train_size <train_set_size> -test_size <test_set_size>"
print "(2) operation name: extract_features"," parameters: -fe_mapper <feature_extraction_mapper> -fe_mapper_params <mapper_params> -fe_reducer <feature_extraction_reducer> -fe_reducer_params <reducer_params> -train_dataset <dataset_location> -train_size <train_set_size> -test_size <test_set_size>"
print "(3) operation name: modeling","\n-->parameters when using single feature set file: -feature_set_location <feature_set_file_location> -model_name <model_name_to_save_as>","\n-->parameters when using directory containing feature set files: -feature_set_location <feature_set_location_directory>"
print "(4) operation name: classification"," parameters: -cl_mapper <classification_mapper> -cl_mapper_params <mapper_params> -cl_reducer <classification_reducer> -cl_reducer_params <reducer_params> -test_dataset <dataset_location> -model <model_file>"
exit()
def run(self):
# print sys.argv
self.arg_obj.parse(sys.argv)
print '*'*100
print "Run got arguments:",self.arg_obj.args
print '*'*100
if not self.arg_obj.args.has_key("operation"):
self.show_help()
if self.arg_obj.args["operation"] == SAMPLING:
self.run_sampling()
if self.arg_obj.args["operation"] == FEATURE_EXTRACTION:
self.run_feature_extraction()
if self.arg_obj.args["operation"] == MODELING:
self.run_modeling()
if self.arg_obj.args["operation"] == CLASSIFICATION:
self.run_classification()
if self.arg_obj.args["operation"] == DEFAULT:
self.run_default_flow()
# method to run sampling
def run_sampling(self):
# launch sampling
_cmd = "python "+_prefix+"/"+"sampler_interface.py " + self.arg_obj.get_string()
print '*'*100
print "run: Invoking Sampling:\n",_cmd
print '*'*100
self.invoke(_cmd)
pass
# method to run feature extraction
def run_feature_extraction(self):
# launch feature extraction here
_cmd = "python "+_prefix+"/"+"feature_extraction_interface.py " + self.arg_obj.get_string()
print "run: Invoking Feature Extraction:\n",_cmd
self.invoke(_cmd)
pass
# method to run modeling
def run_modeling(self):
# launch modeling
_cmd = "python "+_prefix+"/"+"modeler_interface.py " + self.arg_obj.get_string()
print '*'*100
print "run: Invoking Modeling:\n",_cmd
print '*'*100
self.invoke(_cmd)
pass
# method to run the classification
def run_classification(self):
# launch classification
_cmd = "python "+_prefix+"/"+"classification_interface.py " + self.arg_obj.get_string()
print '*'*100
print "run: Invoking Classification:\n",_cmd
print '*'*100
self.invoke(_cmd)
pass
# method to check if all the necessary parameters are provided for default flow
def check_params(self):
# this checks for all params required to execute the default flow of the framework
# mapper and reducer params might be optional- thus they are not required
if not self.arg_obj.args.has_key("fe_mapper") or not self.arg_obj.args.has_key("fe_reducer") or not self.arg_obj.args.has_key("train_dataset") or not self.arg_obj.args.has_key("train_size") or not self.arg_obj.args.has_key("test_size") or not self.arg_obj.args.has_key("cl_mapper") or not self.arg_obj.args.has_key("cl_reducer") :
self.show_help()
# method to run default behavior- here framework handles everything- sampling, feature extraction, modeling
# user just needs to provide required parameters
# this runs: sampling->feature extraction->modeling
def run_default_flow(self):
self.check_params()
# launch feature extraction
self.run_feature_extraction()
# if feature extraction was successful then proceed for modeling
if os.path.exists("./feature_set_for_modeling"):
print "Launching modeler with extracted feature set..."
self.run_modeling()
# if model was successfully generated, set the model parameter for classification flow
self.arg_obj.args["model"] = "trained_models/"+str(self.arg_obj.args["train_size"])+"_output.model"
else:
print "unable to find the directory 'feature_set_for_modeling'"
# if the modeling was successful then proceed for classification
if os.path.exists("./trained_models"):
print "Launching classification with trained model from ./trained_models"
# launch classification
self.run_classification()
else:
print "unable to find the directory 'trained_models'"
# method to invoke the commands
def invoke(self, cmd):
os.system(cmd)
class InterfaceWrapper:
def __init__(self):
self.arg_obj = ArgParser()
# method to show help message
def show_help(self):
print ":( not enough params"
print "usage: python run.py -operation <operation_name> <parameters for operation>"
print "******supported operations******"
print "(1) operation name: sampling", " parameters: -sampler <sampler_implementation> -positive <positive_source_file> -negative <negative_source_file> -train_size <train_set_size> -test_size <test_set_size>"
print "(2) operation name: extract_features", " parameters: -fe_mapper <feature_extraction_mapper> -fe_mapper_params <mapper_params> -fe_reducer <feature_extraction_reducer> -fe_reducer_params <reducer_params> -train_dataset <dataset_location> -train_size <train_set_size> -test_size <test_set_size>"
print "(3) operation name: modeling", "\n-->parameters when using single feature set file: -feature_set_location <feature_set_file_location> -model_name <model_name_to_save_as>", "\n-->parameters when using directory containing feature set files: -feature_set_location <feature_set_location_directory>"
print "(4) operation name: classification", " parameters: -cl_mapper <classification_mapper> -cl_mapper_params <mapper_params> -cl_reducer <classification_reducer> -cl_reducer_params <reducer_params> -test_dataset <dataset_location> -model <model_file>"
exit()
def run(self):
# print sys.argv
self.arg_obj.parse(sys.argv)
print '*' * 100
print "Run got arguments:", self.arg_obj.args
print '*' * 100
if not self.arg_obj.args.has_key("operation"):
self.show_help()
if self.arg_obj.args["operation"] == SAMPLING:
self.run_sampling()
if self.arg_obj.args["operation"] == FEATURE_EXTRACTION:
self.run_feature_extraction()
if self.arg_obj.args["operation"] == MODELING:
self.run_modeling()
if self.arg_obj.args["operation"] == CLASSIFICATION:
self.run_classification()
if self.arg_obj.args["operation"] == DEFAULT:
self.run_default_flow()
# method to run sampling
def run_sampling(self):
# launch sampling
_cmd = "python " + _prefix + "/" + "sampler_interface.py " + self.arg_obj.get_string(
)
print '*' * 100
print "run: Invoking Sampling:\n", _cmd
print '*' * 100
self.invoke(_cmd)
pass
# method to run feature extraction
def run_feature_extraction(self):
# launch feature extraction here
_cmd = "python " + _prefix + "/" + "feature_extraction_interface.py " + self.arg_obj.get_string(
)
print "run: Invoking Feature Extraction:\n", _cmd
self.invoke(_cmd)
pass
# method to run modeling
def run_modeling(self):
# launch modeling
_cmd = "python " + _prefix + "/" + "modeler_interface.py " + self.arg_obj.get_string(
)
print '*' * 100
print "run: Invoking Modeling:\n", _cmd
print '*' * 100
self.invoke(_cmd)
pass
# method to run the classification
def run_classification(self):
# launch classification
_cmd = "python " + _prefix + "/" + "classification_interface.py " + self.arg_obj.get_string(
)
print '*' * 100
print "run: Invoking Classification:\n", _cmd
print '*' * 100
self.invoke(_cmd)
pass
# method to check if all the necessary parameters are provided for default flow
def check_params(self):
# this checks for all params required to execute the default flow of the framework
# mapper and reducer params might be optional- thus they are not required
if not self.arg_obj.args.has_key(
"fe_mapper") or not self.arg_obj.args.has_key(
"fe_reducer") or not self.arg_obj.args.has_key(
"train_dataset") or not self.arg_obj.args.has_key(
"train_size") or not self.arg_obj.args.has_key(
"test_size") or not self.arg_obj.args.has_key(
"cl_mapper"
) or not self.arg_obj.args.has_key(
"cl_reducer"):
self.show_help()
# method to run default behavior- here framework handles everything- sampling, feature extraction, modeling
# user just needs to provide required parameters
# this runs: sampling->feature extraction->modeling
def run_default_flow(self):
self.check_params()
# launch feature extraction
self.run_feature_extraction()
# if feature extraction was successful then proceed for modeling
if os.path.exists("./feature_set_for_modeling"):
print "Launching modeler with extracted feature set..."
self.run_modeling()
# if model was successfully generated, set the model parameter for classification flow
self.arg_obj.args["model"] = "trained_models/" + str(
self.arg_obj.args["train_size"]) + "_output.model"
else:
print "unable to find the directory 'feature_set_for_modeling'"
# if the modeling was successful then proceed for classification
if os.path.exists("./trained_models"):
print "Launching classification with trained model from ./trained_models"
# launch classification
self.run_classification()
else:
print "unable to find the directory 'trained_models'"
# method to invoke the commands
def invoke(self, cmd):
os.system(cmd)