easy-cs-rec-custommodel 0.8.6__py2.py3-none-any.whl

easy_rec/python/test/hive_input_test.py

@@ -0,0 +1,311 @@
+# -*- encoding:utf-8 -*-
+# Copyright (c) Alibaba, Inc. and its affiliates.
+"""Define cv_input, the base class for cv tasks."""
+import logging
+import os
+import unittest
+
+import tensorflow as tf
+from google.protobuf import text_format
+
+from easy_rec.python.input.hive_input import HiveInput
+from easy_rec.python.protos.dataset_pb2 import DatasetConfig
+from easy_rec.python.protos.feature_config_pb2 import FeatureConfig
+from easy_rec.python.protos.hive_config_pb2 import HiveConfig
+from easy_rec.python.protos.pipeline_pb2 import EasyRecConfig
+from easy_rec.python.utils import config_util
+from easy_rec.python.utils import test_utils
+
+if tf.__version__ >= '2.0':
+  import tensorflow.compat.v1 as tf
+
+gfile = tf.gfile
+
+if tf.__version__ >= '2.0':
+  from tensorflow.python.framework.ops import disable_eager_execution
+
+  disable_eager_execution()
+  tf = tf.compat.v1
+
+
+class HiveInputTest(tf.test.TestCase):
+
+  def _init_config(self):
+    hive_host = os.environ['hive_host']
+    hive_username = os.environ['hive_username']
+    hive_table_name = os.environ['hive_table_name']
+    hive_hash_fields = os.environ['hive_hash_fields']
+
+    hive_train_input = """
+      host: "{}"
+      username: "{}"
+      table_name: "{}"
+      limit_num: 500
+      hash_fields: "{}"
+    """.format(hive_host, hive_username, hive_table_name, hive_hash_fields)
+    hive_eval_input = """
+      host: "{}"
+      username: "{}"
+      table_name: "{}"
+      limit_num: 500
+      hash_fields: "{}"
+    """.format(hive_host, hive_username, hive_table_name, hive_hash_fields)
+    self.hive_train_input_config = HiveConfig()
+    text_format.Merge(hive_train_input, self.hive_train_input_config)
+
+    self.hive_eval_input_config = HiveConfig()
+    text_format.Merge(hive_eval_input, self.hive_eval_input_config)
+
+  def __init__(self, methodName='HiveInputTest'):
+    super(HiveInputTest, self).__init__(methodName=methodName)
+
+  @unittest.skipIf('hive_host' not in os.environ or
+                   'hive_username' not in os.environ or
+                   'hive_table_name' not in os.environ or
+                   'hive_hash_fields' not in os.environ,
+                   """Only execute hive_config var are specified,hive_host、
+       hive_username、hive_table_name、hive_hash_fields is available.""")
+  def test_hive_input(self):
+    self._init_config()
+    data_config_str = """
+          batch_size: 1024
+          label_fields: "label_1"
+          label_fields: "label_2"
+          num_epochs: 1
+          prefetch_size: 32
+          input_type: HiveInput
+          input_fields {
+            input_name:'label_1'
+            input_type: INT32
+          }
+          input_fields {
+            input_name:'label_2'
+            input_type: INT32
+          }
+          input_fields {
+            input_name:'age'
+            input_type: INT32
+          }
+          input_fields {
+            input_name: "class_of_worker"
+          }
+          input_fields {
+            input_name: "industry_code"
+          }
+          input_fields {
+            input_name: "occupation_code"
+          }
+          input_fields {
+            input_name: "education"
+          }
+          input_fields {
+            input_name: "wage_per_hour"
+            input_type: DOUBLE
+          }
+          input_fields {
+            input_name: "enrolled_in_edu_inst_last_wk"
+          }
+          input_fields {
+            input_name: "major_industry"
+          }
+          input_fields {
+            input_name: "major_occupation"
+          }
+          input_fields {
+            input_name: "mace"
+          }
+          input_fields {
+            input_name: "hispanic_origin"
+          }
+          input_fields {
+            input_name: "sex"
+          }
+          input_fields {
+            input_name: "member_of_a_labor_union"
+          }
+          input_fields {
+            input_name: "reason_for_unemployment"
+          }
+          input_fields {
+            input_name: "full_or_part_time_employment_stat"
+          }
+          input_fields {
+            input_name: "capital_gains"
+            input_type: DOUBLE
+          }
+          input_fields {
+            input_name: "capital_losses"
+            input_type: DOUBLE
+          }
+          input_fields {
+            input_name: "divdends_from_stocks"
+            input_type: DOUBLE
+          }
+          input_fields {
+            input_name: "tax_filer_status"
+          }
+          input_fields {
+            input_name: "region_of_previous_residence"
+          }
+          input_fields {
+            input_name: "state_of_previous_residence"
+          }
+          input_fields {
+            input_name: "detailed_household_and_family_stat"
+          }
+          input_fields {
+            input_name: "detailed_household_summary_in_household"
+          }
+          input_fields {
+            input_name: "instance_weight"
+          }
+          input_fields {
+            input_name: "migration_code_change_in_msa"
+          }
+          input_fields {
+            input_name: "migration_code_change_in_reg"
+          }
+          input_fields {
+            input_name: "migration_code_move_within_reg"
+          }
+          input_fields {
+            input_name: "live_in_this_house_1_year_ago"
+          }
+          input_fields {
+            input_name: "migration_prev_res_in_sunbelt"
+          }
+          input_fields {
+            input_name: "num_persons_worked_for_employer"
+            input_type: INT32
+          }
+          input_fields {
+            input_name: "family_members_under_18"
+          }
+          input_fields {
+            input_name: "country_of_birth_father"
+          }
+          input_fields {
+            input_name: "country_of_birth_mother"
+          }
+          input_fields {
+            input_name: "country_of_birth_self"
+          }
+          input_fields {
+            input_name: "citizenship"
+          }
+          input_fields {
+            input_name: "own_business_or_self_employed"
+          }
+          input_fields {
+            input_name: "fill_inc_questionnaire_for_veteran_s_admin"
+          }
+          input_fields {
+            input_name: "veterans_benefits"
+          }
+          input_fields {
+            input_name: "weeks_worked_in_year"
+            input_type: INT32
+          }
+          input_fields {
+            input_name: "year"
+          }
+      """
+
+    feature_config_str = """
+      input_names: "own_business_or_self_employed"
+      feature_type: IdFeature
+      embedding_dim: 9
+      hash_bucket_size: 400
+      embedding_name: "feature"
+      """
+
+    dataset_config = DatasetConfig()
+    text_format.Merge(data_config_str, dataset_config)
+    feature_config = FeatureConfig()
+    text_format.Merge(feature_config_str, feature_config)
+    feature_configs = [feature_config]
+
+    empty_config = FeatureConfig()
+    empty_config.CopyFrom(feature_config)
+    while len(empty_config.input_names) > 0:
+      empty_config.input_names.pop()
+    while len(empty_config.shared_names) > 0:
+      empty_config.shared_names.pop()
+    train_input_fn = HiveInput(dataset_config, feature_configs,
+                               self.hive_train_input_config).create_input()
+    dataset = train_input_fn(mode=tf.estimator.ModeKeys.TRAIN)
+    iterator = dataset.make_initializable_iterator()
+    tf.add_to_collection(tf.GraphKeys.TABLE_INITIALIZERS, iterator.initializer)
+    features, labels = iterator.get_next()
+    init_op = tf.get_collection(tf.GraphKeys.TABLE_INITIALIZERS)
+    gpu_options = tf.GPUOptions(allow_growth=True)
+    session_config = tf.ConfigProto(
+        gpu_options=gpu_options,
+        allow_soft_placement=True,
+        log_device_placement=False)
+    with self.test_session(config=session_config) as sess:
+      sess.run(init_op)
+      feature_dict, label_dict = sess.run([features, labels])
+      for key in feature_dict:
+        print(key, feature_dict[key][:5])
+
+      for key in label_dict:
+        print(key, label_dict[key][:5])
+    return 0
+
+  @unittest.skipIf('hive_host' not in os.environ or
+                   'hive_username' not in os.environ or
+                   'hive_table_name' not in os.environ or
+                   'hive_hash_fields' not in os.environ,
+                   """Only execute hive_config var are specified,hive_host、
+       hive_username、hive_table_name、hive_hash_fields is available.""")
+  def test_mmoe(self):
+    pipeline_config_path = 'samples/emr_script/mmoe/mmoe_census_income.config'
+    gpus = test_utils.get_available_gpus()
+    if len(gpus) > 0:
+      test_utils.set_gpu_id(gpus[0])
+    else:
+      test_utils.set_gpu_id(None)
+
+    if not isinstance(pipeline_config_path, EasyRecConfig):
+      logging.info('testing pipeline config %s' % pipeline_config_path)
+    if 'TF_CONFIG' in os.environ:
+      del os.environ['TF_CONFIG']
+
+    if isinstance(pipeline_config_path, EasyRecConfig):
+      pipeline_config = pipeline_config_path
+    else:
+      pipeline_config = test_utils._load_config_for_test(
+          pipeline_config_path, self._test_dir)
+
+    pipeline_config.train_config.train_distribute = 0
+    pipeline_config.train_config.num_gpus_per_worker = 1
+    pipeline_config.train_config.sync_replicas = False
+
+    config_util.save_pipeline_config(pipeline_config, self._test_dir)
+    test_pipeline_config_path = os.path.join(self._test_dir, 'pipeline.config')
+    hyperparam_str = ''
+    train_cmd = 'python -m easy_rec.python.train_eval --pipeline_config_path %s %s' % (
+        test_pipeline_config_path, hyperparam_str)
+    proc = test_utils.run_cmd(train_cmd,
+                              '%s/log_%s.txt' % (self._test_dir, 'master'))
+    proc.wait()
+    if proc.returncode != 0:
+      logging.error('train %s failed' % test_pipeline_config_path)
+      return 1
+    return 0
+
+  def setUp(self):
+    logging.info('Testing %s.%s' % (type(self).__name__, self._testMethodName))
+    self._test_dir = test_utils.get_tmp_dir()
+    self._success = True
+    logging.info('test dir: %s' % self._test_dir)
+
+  def tearDown(self):
+    test_utils.set_gpu_id(None)
+    if self._success:
+      test_utils.clean_up(self._test_dir)
+
+
+if __name__ == '__main__':
+  tf.test.main()

easy_rec/python/test/hpo_test.py

@@ -0,0 +1,235 @@
+# -*- encoding:utf-8 -*-
+# Copyright (c) Alibaba, Inc. and its affiliates.
+
+import json
+import logging
+import os
+import time
+
+import numpy as np
+import tensorflow as tf
+
+from easy_rec.python.protos.feature_config_pb2 import FeatureConfig
+from easy_rec.python.utils import config_util
+from easy_rec.python.utils import hpo_util
+from easy_rec.python.utils import test_utils
+
+if tf.__version__ >= '2.0':
+  import tensorflow.io.gfile as gfile
+  from tensorflow.core.protobuf import config_pb2
+
+  ConfigProto = config_pb2.ConfigProto
+  GPUOptions = config_pb2.GPUOptions
+else:
+  from tensorflow.python.platform import gfile
+  GPUOptions = tf.GPUOptions
+  ConfigProto = tf.ConfigProto
+
+
+class HPOTest(tf.test.TestCase):
+
+  def __init__(self, methodName='HPOTest'):
+    super(HPOTest, self).__init__(methodName=methodName)
+    self._metric_data_path = 'data/test/hpo_test/eval_val/*.tfevents.*'
+
+  def test_get_metric(self):
+    vals = hpo_util.get_all_eval_result(self._metric_data_path)
+    logging.info('eval result num = %d' % len(vals))
+    logging.info('eval result[0] = %s' % json.dumps(vals[0]))
+
+  def load_config(self, config_path):
+    with gfile.GFile(config_path, 'r') as fin:
+      return json.load(fin)['param']
+
+  def test_save_eval_metrics(self):
+    test_dir = test_utils.get_tmp_dir()
+    tmp_file = os.path.join(test_dir,
+                            'easy_rec_hpo_test_%d.metric' % time.time())
+    hpo_util.save_eval_metrics('data/test/hpo_test/', tmp_file, False)
+    test_utils.clean_up(test_dir)
+
+  def test_edit_config(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    assert tmp_config.feature_config.features[0].embedding_dim == 120
+
+  def test_edit_config_v2(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v2.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for tmp_fea in tmp_config.feature_configs:
+      if tmp_fea.input_names[0] == 'site_id':
+        assert tmp_fea.embedding_dim == 32
+      else:
+        assert tmp_fea.embedding_dim == 16
+
+  def test_edit_config_v3(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v3.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if i >= 10 and i < 20:
+        assert tmp_fea.embedding_dim == 37
+      else:
+        assert tmp_fea.embedding_dim == 16
+
+  def test_edit_config_v4(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v4.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if i < 15:
+        assert tmp_fea.embedding_dim == 37
+      else:
+        assert tmp_fea.embedding_dim == 16
+
+  def test_edit_config_v5(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v5.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if i >= 5:
+        assert tmp_fea.embedding_dim == 37
+      else:
+        assert tmp_fea.embedding_dim == 16
+
+  def test_edit_config_v51(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v51.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if i == 5:
+        assert tmp_fea.embedding_dim == 37
+
+  def test_edit_config_v6(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v6.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] >= 'site':
+        assert tmp_fea.embedding_dim == 32, 'input_name = %s %d' % (
+            tmp_fea.input_names[0], tmp_fea.embedding_dim)
+      else:
+        assert tmp_fea.embedding_dim == 16
+
+  def test_edit_config_v7(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v7.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'c21':
+        assert len(tmp_fea.boundaries) == 4 and np.abs(tmp_fea.boundaries[0] -
+                                                       10.0) < 1e-5
+
+  def test_edit_config_v71(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v71.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'c21':
+        assert len(tmp_fea.boundaries) == 4 and np.abs(tmp_fea.boundaries[0] -
+                                                       10.0) < 1e-5
+
+  def test_edit_config_v8(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v8.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'c21':
+        assert len(tmp_fea.boundaries) == 4 and np.abs(tmp_fea.boundaries[0] -
+                                                       4.0) < 1e-5
+        assert tmp_fea.embedding_dim == 32
+
+  def test_edit_config_v81(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v81.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.feature_type == tmp_fea.RawFeature:
+        assert tmp_fea.embedding_dim == 24
+
+  def test_edit_config_v9(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v9.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    assert tmp_config.train_config.fine_tune_checkpoint == \
+           'oss://easy-rec/test/experiment/ctr_v93/model.ckpt-1000'
+
+  def test_edit_config_v10(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v10.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'c21':
+        assert len(tmp_fea.boundaries) == 4 and np.abs(tmp_fea.boundaries[0] -
+                                                       4.0) < 1e-5
+        assert tmp_fea.embedding_dim == 32
+
+  def test_edit_config_v11(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v11.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'c21':
+        assert len(tmp_fea.boundaries) == 4 and np.abs(tmp_fea.boundaries[0] -
+                                                       10.0) < 1e-5
+
+  def test_edit_config_v12(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v12.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'c21':
+        assert len(tmp_fea.boundaries) == 25
+        assert np.abs(tmp_fea.boundaries[1] - 21.0) < 1e-5
+
+  def test_edit_config_v13(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v13.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    assert not tmp_config.export_config.multi_placeholder
+
+  def test_edit_config_v14(self):
+    tmp_file = 'samples/model_config/deepfm_multi_cls_on_avazu_ctr.config'
+    tmp_config = config_util.get_configs_from_pipeline_file(tmp_file)
+    tmp_file = 'samples/hpo/hpo_param_v14.json'
+    tmp_config = config_util.edit_config(tmp_config, self.load_config(tmp_file))
+    for i, tmp_fea in enumerate(tmp_config.feature_configs):
+      if tmp_fea.input_names[0] == 'hour':
+        assert len(tmp_fea.feature_type) == FeatureConfig.RawFeature
+
+  def test_save_eval_metrics_with_env(self):
+    os.environ['TF_CONFIG'] = """
+                              { "cluster": {
+                                  "worker": ["127.0.0.1:2020"],
+                                  "chief": ["127.0.0.1:2021"]
+                                 },
+                                 "task": {"type": "chief", "index": 0}
+                              }
+                              """
+    test_dir = test_utils.get_tmp_dir()
+    tmp_file = os.path.join(test_dir,
+                            'easy_rec_hpo_test_%d.metric' % time.time())
+    hpo_util.save_eval_metrics('data/test/hpo_test/', tmp_file, False)
+    test_utils.clean_up(test_dir)
+
+
+if __name__ == '__main__':
+  tf.test.main()