def start():
sconf = SparkConf()
sconf.set('spark.cores.max', 3)
sc = SparkContext(appName='pysparkdfka', conf=sconf)
ssc = StreamingContext(sc, 5)
brokers = "10.12.64.205:9092"
topic = 'test'
stream = KafkaUtils.createDirectStream(
ssc, [topic], kafkaParams={"metadata.broker.list": brokers})
json = stream.map(lambda v: v[1])
json.foreachRDD(lambda rdd: rdd.foreach(echo))
json.pprint()
ssc.start()
ssc.awaitTermination()
def _cli_eval(config, output_dir, args):
# Load model config from previous experiment
with open(os.path.join(output_dir, 'config.yml'), 'r') as f:
model_config = yaml.load(f)['model']
model_config.update(config.get('model', {}))
config['model'] = model_config
results = evaluate(config, output_dir, n_iter=config.get('eval_iter'))
# Print and export results
logging.info('Evaluation results: \n{}'.format(
pprint(results, indent=2, default=str)))
with open(os.path.join(output_dir, 'eval.txt'), 'a') as f:
f.write('Evaluation for {} dataset:\n'.format(config['data']['name']))
for r, v in results.items():
f.write('\t{}:\n\t\t{}\n'.format(r, v))
f.write('\n')