def create_recommendation_microservice(self,
recommender_folder,
memcache_servers=None,
memcache_pool_size=2):
"""
create recommedation Flask microservice app
Parameters
----------
recommender_folder : str
location of recommender model files
memcache_servers : comma separated string, optional
memcache server locations, e.g., 127.0.0.1:11211
memcache_pool_size : int, optional
size of memcache pool
"""
app = Flask(__name__)
if not memcache_servers is None:
mc = pylibmc.Client(memcache_servers)
_mc_pool = pylibmc.ClientPool(mc, memcache_pool_size)
app.config["seldon_memcache"] = _mc_pool
rw = seldon.Recommender_wrapper()
recommender = rw.load_recommender(recommender_folder)
app.config["seldon_recommender"] = recommender
app.register_blueprint(recommend_blueprint)
# other setup tasks
return app
import sys, getopt, argparse
from seldon.text.ngram_recommend import NgramModel
if __name__ == '__main__':
parser = argparse.ArgumentParser(prog='upload model')
parser.add_argument('--arpa', help='arpa file', required=True)
parser.add_argument('--dst', help='dst folder', required=True)
parser.add_argument(
'--aws_key',
help='aws key - needed if input or output is on AWS and no IAM')
parser.add_argument(
'--aws_secret',
help='aws secret - needed if input or output on AWS and no IAM')
args = parser.parse_args()
opts = vars(args)
recommender = NgramModel()
recommender.fit(args.arpa)
import seldon
if "aws_key" in opts:
rw = seldon.Recommender_wrapper(aws_key=args.aws_key,
aws_secret=args.aws_secret)
else:
rw = seldon.Recommender_wrapper()
rw.save_recommender(recommender, args.dst)
docs = []
for filename in os.listdir("reuters-21578-json/data/full"):
f = open("reuters-21578-json/data/full/" + filename)
js = json.load(f)
for j in js:
if 'topics' in j and 'body' in j:
d = {}
d["id"] = j['id']
d["text"] = j['body'].replace("\n", "")
d["title"] = j['title']
d["tags"] = ",".join(j['topics'])
docs.append(d)
print "loaded ", len(docs), " documents"
from seldon.text import DocumentSimilarity, DefaultJsonCorpus
import logging
logger = logging.getLogger()
logger.setLevel(logging.INFO)
corpus = DefaultJsonCorpus(docs)
ds = DocumentSimilarity(model_type='gensim_lsi')
ds.fit(corpus)
print "built model"
import seldon
rw = seldon.Recommender_wrapper()
rw.save_recommender(ds, "reuters_recommender")
print "saved recommender"