model: tensorflow: Back out tensorflow 2.X support temporarily

Related: #374

Signed-off-by: John Andersen <[email protected]>

Author: pdxjohnny

CHANGELOG.md

@@ -40,7 +40,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   to setuptools `setup` function within a `setup.py` file.
 ### Changed
 - All instances of `src_url` changed to `key`.
-- Moved from tensorflow 1 to tensorflow 2.
 - `readonly` parameter in source config is now changed to `readwrite`.
 - `predict` parameter of all model config classes has been changed from `str` to `Feature`.
 - Defining features on the command line no longer requires that defined features

model/tensorflow/dffml_model_tensorflow/dnnc.py

@@ -10,9 +10,7 @@
 from typing import List, Dict, Any, AsyncIterator, Tuple, Optional, Type
 
 import numpy as np
-
-os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
-import tensorflow as tf
+import tensorflow
 
 from dffml.repo import Repo
 from dffml.feature import Feature, Features
@@ -67,7 +65,7 @@ def _feature_feature_column(self, feature: Feature):
             or dtype is float
             or issubclass(dtype, float)
         ):
-            return tf.feature_column.numeric_column(
+            return tensorflow.feature_column.numeric_column(
                 feature.NAME, shape=feature.length()
             )
         self.logger.warning(
@@ -114,7 +112,7 @@ async def predict_input_fn(self, repos: AsyncIterator[Repo], **kwargs):
         self.logger.info("------ Repo Data ------")
         self.logger.info("x_cols:    %d", len(list(x_cols.values())[0]))
         self.logger.info("-----------------------")
-        input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
+        input_fn = tensorflow.estimator.inputs.numpy_input_fn(
             x_cols, shuffle=False, num_epochs=1, **kwargs
         )
         return input_fn, ret_repos
@@ -208,7 +206,7 @@ def model(self):
             len(self.classifications),
             self.classifications,
         )
-        self._model = tf.estimator.DNNClassifier(
+        self._model = tensorflow.estimator.DNNClassifier(
             feature_columns=list(self.feature_columns.values()),
             hidden_units=self.parent.config.hidden,
             n_classes=len(self.parent.config.classifications),
@@ -249,7 +247,7 @@ async def training_input_fn(
         self.logger.info("x_cols:    %d", len(list(x_cols.values())[0]))
         self.logger.info("y_cols:    %d", len(y_cols))
         self.logger.info("-----------------------")
-        input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
+        input_fn = tensorflow.estimator.inputs.numpy_input_fn(
             x_cols,
             y_cols,
             batch_size=self.parent.config.batchsize,
@@ -289,7 +287,7 @@ async def accuracy_input_fn(
         self.logger.info("x_cols:    %d", len(list(x_cols.values())[0]))
         self.logger.info("y_cols:    %d", len(y_cols))
         self.logger.info("-----------------------")
-        input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
+        input_fn = tensorflow.estimator.inputs.numpy_input_fn(
             x_cols,
             y_cols,
             batch_size=self.parent.config.batchsize,

model/tensorflow/dffml_model_tensorflow/dnnr.py

@@ -6,21 +6,19 @@
 from typing import List, Dict, Any, AsyncIterator
 
 import numpy as np
-
-os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
-import tensorflow as tf
+import tensorflow
 
 from dffml.repo import Repo
-from dffml.util.cli.arg import Arg
+from dffml.source.source import Sources
 from dffml.model.model import Model
 from dffml.accuracy import Accuracy
-from dffml.source.source import Sources
 from dffml.util.entrypoint import entrypoint
-from dffml.util.cli.parser import list_action
 from dffml.base import BaseConfig, config, field
+from dffml.util.cli.arg import Arg
 from dffml.feature.feature import Feature, Features
+from dffml.util.cli.parser import list_action
 
-from .dnnc import TensorflowModelContext
+from dffml_model_tensorflow.dnnc import TensorflowModelContext
 
 
 @config
@@ -65,8 +63,9 @@ def model(self):
         if self._model is not None:
             return self._model
         self.logger.debug("Loading model ")
-        _head = tf.estimator.RegressionHead()
-        self._model = tf.estimator.DNNEstimator(
+
+        _head = tensorflow.contrib.estimator.regression_head()
+        self._model = tensorflow.estimator.DNNEstimator(
             head=_head,
             feature_columns=list(self.feature_columns.values()),
             hidden_units=self.parent.config.hidden,
@@ -103,7 +102,7 @@ async def training_input_fn(
         self.logger.info("x_cols:    %d", len(list(x_cols.values())[0]))
         self.logger.info("y_cols:    %d", len(y_cols))
         self.logger.info("-----------------------")
-        input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
+        input_fn = tensorflow.estimator.inputs.numpy_input_fn(
             x_cols,
             y_cols,
             batch_size=batch_size,
@@ -139,7 +138,7 @@ async def evaluate_input_fn(
         self.logger.info("x_cols:    %d", len(list(x_cols.values())[0]))
         self.logger.info("y_cols:    %d", len(y_cols))
         self.logger.info("-----------------------")
-        input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
+        input_fn = tensorflow.estimator.inputs.numpy_input_fn(
             x_cols,
             y_cols,
             batch_size=batch_size,

model/tensorflow/setup.py

@@ -18,7 +18,7 @@
 with open(os.path.join(self_path, "README.md"), "r", encoding="utf-8") as f:
     readme = f.read()
 
-INSTALL_REQUIRES = ["tensorflow>=2.1.0rc2"] + (
+INSTALL_REQUIRES = ["tensorflow==1.14.0"] + (
     ["dffml>=0.3.1"]
     if not any(
         list(

model/tensorflow/tests/test_dnnc.py

@@ -55,7 +55,7 @@ def setUpClass(cls):
                 directory=cls.model_dir.name,
                 steps=1000,
                 epochs=30,
-                hidden=[300, 200, 80, 10],
+                hidden=[10, 20, 10],
                 predict=DefFeature("string", str, 1),
                 classifications=["a", "not a"],
                 clstype=str,

model/tensorflow/tests/test_dnnr.py

@@ -51,7 +51,7 @@ def setUpClass(cls):
                 directory=cls.model_dir.name,
                 steps=1000,
                 epochs=30,
-                hidden=[200, 100, 80, 10],
+                hidden=[10, 20, 10],
                 predict=DefFeature("TARGET", float, 1),
                 features=cls.features,
             )