remove unused layers and switch some ops from tf to keras

.github/workflows/ci_tests.yml

@@ -29,7 +29,6 @@ jobs:
           pip install torch~=${{ matrix.TORCH_VER }}
           pip install keras coveralls pydot graphviz pytest pytest-cov
           pip install .
-          python -c "import keras; print('Keras Version: ', keras.__version__)"
       - name: Cache test data
         uses: actions/cache@v3
         with:

astroNN/models/__init__.py

@@ -141,7 +141,7 @@ def load_folder(folder=None):
         # else try to import it from standard way
         try:
             astronn_model_obj = getattr(
-                importlib.import_module(f"astroNN.models"), identifier
+                importlib.import_module("astroNN.models"), identifier
             )()
         except ImportError:
             # try to load custom model from CUSTOM_MODEL_PATH if none are working
@@ -186,9 +186,9 @@ def load_folder(folder=None):
     # Must have parameter
     astronn_model_obj._input_shape = parameter["input"]
     astronn_model_obj._labels_shape = parameter["labels"]
-    if type(astronn_model_obj._input_shape) is not dict:
+    if not isinstance(astronn_model_obj._input_shape, dict):
         astronn_model_obj._input_shape = {"input": astronn_model_obj._input_shape}
-    if type(astronn_model_obj._labels_shape) is not dict:
+    if not isinstance(astronn_model_obj._labels_shape, dict):
         astronn_model_obj._labels_shape = {"output": astronn_model_obj._labels_shape}
     astronn_model_obj.num_hidden = parameter["hidden"]
     astronn_model_obj.input_norm_mode = parameter["input_norm_mode"]
@@ -313,16 +313,10 @@ def load_folder(folder=None):
         # its weird that keras needs -> metrics[metric][0] instead of metrics[metric] likes losses
         try:
             try:
-                if version.parse(tf.__version__) >= version.parse("2.4.0"):
-                    metrics = [
-                        losses_lookup(_metric["config"]["fn"])
-                        for _metric in metrics_raw[0]
-                    ]
-                else:
-                    metrics = [
-                        losses_lookup(metrics_raw[_metric][0])
-                        for _metric in metrics_raw
-                    ]
+                metrics = [
+                    losses_lookup(_metric["config"]["fn"])
+                    for _metric in metrics_raw[0]
+                ]
             except TypeError:
                 metrics = [losses_lookup(metrics_raw[0])]
         except:

astroNN/models/misc_models.py

@@ -5,7 +5,7 @@
 
 from astroNN.models.base_bayesian_cnn import BayesianCNNBase
 from astroNN.models.base_cnn import CNNBase
-from astroNN.nn.layers import MCDropout, PolyFit
+from astroNN.nn.layers import MCDropout
 from astroNN.nn.losses import (
     bayesian_binary_crossentropy_wrapper,
     bayesian_binary_crossentropy_var_wrapper,
@@ -257,47 +257,3 @@ def model(self):
             )
 
         return model, model_prediction, output_loss, variance_loss
-
-
-# noinspection PyCallingNonCallable
-class SimplePolyNN(CNNBase):
-    """
-    Class for Neural Network for Gaia Polynomial fitting
-
-    :History: 2018-Jul-23 - Written - Henry Leung (University of Toronto)
-    """
-
-    def __init__(self, lr=0.005, init_w=None, use_xbias=False):
-        super().__init__()
-
-        self._implementation_version = "1.0"
-        self.max_epochs = 40
-        self.lr = lr
-        self.reduce_lr_epsilon = 0.00005
-        self.num_hidden = 3  # equals degree of polynomial to fit
-
-        self.reduce_lr_min = 1e-8
-        self.reduce_lr_patience = 2
-
-        self.input_norm_mode = 0
-        self.labels_norm_mode = 0
-        self.init_w = init_w
-        self.use_xbias = use_xbias
-        self.task = "regression"
-        self.targetname = ["unbiased_parallax"]
-
-    def model(self):
-        input_tensor = Input(shape=self._input_shape, name="input")
-        flattener = Flatten()(input_tensor)
-        output = PolyFit(
-            deg=self.num_hidden,
-            output_units=self._labels_shape,
-            use_xbias=self.use_xbias,
-            name="output",
-            init_w=self.init_w,
-            kernel_regularizer=regularizers.l2(self.l2),
-        )(flattener)
-
-        model = Model(inputs=input_tensor, outputs=output)
-
-        return model

astroNN/nn/__init__.py

@@ -1,45 +1,3 @@
-def reduce_var(x, axis=None, keepdims=False):
-    """
-    Calculate variance using Tensorflow (as opposed to tf.nn.moment which return both variance and mean)
-
-    :param x: Data
-    :type x: tf.Tensor
-    :param axis: Axis
-    :type axis: int
-    :param keepdims: Keeping variance dimension as data or not
-    :type keepdims: boolean
-    :return: Variance
-    :rtype: tf.Tensor
-    :History: 2018-Mar-04 - Written - Henry Leung (University of Toronto)
-    """
-    import tensorflow as tf
-
-    m = tf.reduce_mean(x, axis, True)
-    devs_squared = tf.square(x - m)
-    return tf.reduce_mean(devs_squared, axis, keepdims)
-
-
-def intpow_avx2(x, n):
-    """
-    Calculate integer power of float (including negative) even with Tensorflow compiled with AVX2 since --fast-math
-    compiler flag aggressively optimize float operation which is common with AVX2 flag
-
-    :param x: identifier
-    :type x: tf.Tensor
-    :param n: an integer power (a float will be casted to integer!!)
-    :type n: int
-    :return: powered float(s)
-    :rtype: tf.Tensor
-    :History: 2018-Aug-13 - Written - Henry Leung (University of Toronto)
-    """
-    import tensorflow as tf
-
-    # expand inputs to prepare to be tiled
-    expanded_inputs = tf.expand_dims(x, 1)
-    # we want [1, self.n]
-    return tf.reduce_prod(tf.tile(expanded_inputs, [1, n]), axis=-1)
-
-
 def nn_obj_lookup(identifier, module_obj=None, module_name="default_obj"):
     """
     Lookup astroNN.nn function by name