模型调试

train/down_vocab.py

@@ -10,7 +10,10 @@ def getRedisConn():
     conn = redis.Redis(host="172.18.51.10", port=6379, db=0, decode_responses = True) #test
     return conn
 
-save_dir = sys.argv[1]
+if len(sys.argv) == 2:
+    save_dir = sys.argv[1]
+else:
+    save_dir = '/data/files/wideAndDeep/'
 
 print('save_dir: ', save_dir)
 
@@ -34,3 +37,6 @@ for vocab_key in vocab_keys:
         print('texts: ', len(texts))
         f.write('\n'.join(texts))
 
+os.system("hdfs dfs -getmerge /strategy/train_samples {save_dir}train_samples.csv".format(save_dir = save_dir))
+os.system("hdfs dfs -getmerge /strategy/test_samples {save_dir}test_samples.csv".format(save_dir = save_dir))
+

train/train_service_sk.py

@@ -2,7 +2,7 @@ import tensorflow as tf
 tf.logging.set_verbosity(tf.logging.INFO)
 import sys
 
-BASE_DIR = '/Users/edz/software/Recommend/'
+BASE_DIR = '/data/files/wideAndDeep/'
 
 def input_fn(csv_path, epoch, shuffle, batch_size):
     dataset = tf.data.TextLineDataset(csv_path)
@@ -211,7 +211,7 @@ session_config = tf.compat.v1.ConfigProto(log_device_placement = True, allow_sof
 
 config = tf.estimator.RunConfig(save_checkpoints_steps = 10000)
 
-wideAndDeepModel = tf.estimator.DNNLinearCombinedClassifier(model_dir = './wideAndDeep_tf1',
+wideAndDeepModel = tf.estimator.DNNLinearCombinedClassifier(model_dir = BASE_DIR + 'model',
     linear_feature_columns = linear_feature_columns,
     dnn_feature_columns = dnn_feature_columns,
     dnn_hidden_units = [128, 32],