Merge pull request #2076 from GPflow/develop

Release 2.8.1
GPflow · Jun 27, 2023 · 8549323 · 8549323
2 parents 024d381 a602aff
commit 8549323
Show file tree

Hide file tree

Showing 10 changed files with 46 additions and 16 deletions.
diff --git a/.circleci/config.yml b/.circleci/config.yml
@@ -12,13  12,13 @@ parameters:
     default: "2.4.0"
   max_tf_ver:
     type: string
-    default: "2.10.0"
     default: "2.12.0"
   min_tfp_ver:
     type: string
     default: "0.12.0"
   max_tfp_ver:
     type: string
-    default: "0.18.0"
     default: "0.20.0"
   min_venv_dir:
     type: string
     default: min_venv

diff --git a/.gitignore b/.gitignore
@@ -77,3  77,6 @@ target/
 
 # TensorFlow logs
 events.out.tfevents.*
 
 # pyenv
 .python-version
diff --git a/CONTRIBUTORS.md b/CONTRIBUTORS.md
@@ -88,5  88,6 @@ Because GitHub's [graph of contributors](http://github.com/GPflow/GPflow/graphs/
 [@sethaxen](https://github.com/sethaxen)
 [@khurram-ghani](https://github.com/khurram-ghani)
 [@partev](https://github.com/partev)
 [@uri-granta](https://github.com/uri-granta)
 
 Add yourself when you first contribute to GPflow's code, tests, or documentation!
diff --git a/RELEASE.md b/RELEASE.md
@@ -66,6  66,21 @@ This release contains contributions from:
 <INSERT>, <NAME>, <HERE>, <USING>, <GITHUB>, <HANDLE>
 
 
 # Release 2.8.1
 
 A small fix to ensure support for (and testing with) TensorFlow 2.12.
 
 ## Bug Fixes and Other Changes
 
 * Support and test with TensorFlow 2.12
 
 ## Thanks to our Contributors
 
 This release contains contributions from:
 
 uri-granta
 
 
 # Release 2.8.0
 
 The main focus of this release is to provide users control over arguments for `tf.function`

diff --git a/VERSION b/VERSION
@@ -1  1 @@
-2.8.0
 2.8.1
diff --git a/doc/sphinx/conf.py b/doc/sphinx/conf.py
@@ -24,7  24,7 @@
 author = "The GPflow Contributors"
 
 # The full version, including alpha/beta/rc tags
-release = "2.8.0"
 release = "2.8.1"
 
 # -- General configuration ---------------------------------------------------
 

diff --git a/doc/sphinx/installation.rst b/doc/sphinx/installation.rst
@@ -25,6  25,10 @@ explicitly install specific versions of these.
  --------------------- --------------------------------- 
 | 2.10.*              | 0.18.*                          |
  --------------------- --------------------------------- 
 | 2.11.*              | 0.19.*                          |
  --------------------- --------------------------------- 
 | 2.12.*              | 0.20.*                          |
  --------------------- --------------------------------- 
 
 Second, a word of warning about new Mac computers. On new Mac machines you will need to install
 ``tensorflow-macos`` instead of the regular ``tensorflow``.

diff --git a/doc/sphinx/notebooks/advanced/natural_gradients.pct.py b/doc/sphinx/notebooks/advanced/natural_gradients.pct.py
@@ -22,6  22,12 @@
 import numpy as np
 import tensorflow as tf
 
 try:
     # use legacy Adam optimizer to support old TF versions
     from tensorflow.keras.optimizers.legacy import Adam
 except ImportError:
     from tensorflow.keras.optimizers import Adam
 
 import gpflow
 from gpflow import set_trainable
 from gpflow.ci_utils import reduce_in_tests
@@ -115,8  121,8 @@
 set_trainable(vgp.q_mu, False)
 set_trainable(vgp.q_sqrt, False)
 
-adam_opt_for_vgp = tf.optimizers.Adam(adam_learning_rate)
-adam_opt_for_gpr = tf.optimizers.Adam(adam_learning_rate)
 adam_opt_for_vgp = Adam(adam_learning_rate)
 adam_opt_for_gpr = Adam(adam_learning_rate)
 
 # %%
 for i in range(iterations):
@@ -243,15  249,15 @@
 )
 
 # ordinary gradients with Adam for SVGP
-ordinary_adam_opt = tf.optimizers.Adam(adam_learning_rate)
 ordinary_adam_opt = Adam(adam_learning_rate)
 
 # NatGrads and Adam for SVGP
 # Stop Adam from optimizing the variational parameters
 set_trainable(svgp_natgrad.q_mu, False)
 set_trainable(svgp_natgrad.q_sqrt, False)
 
 # Create the optimize_tensors for SVGP
-natgrad_adam_opt = tf.optimizers.Adam(adam_learning_rate)
 natgrad_adam_opt = Adam(adam_learning_rate)
 
 natgrad_opt = NaturalGradient(gamma=0.1)
 variational_params = [(svgp_natgrad.q_mu, svgp_natgrad.q_sqrt)]
@@ -333,15  339,15 @@
 )
 
 # ordinary gradients with Adam for VGP with Bernoulli likelihood
-adam_opt = tf.optimizers.Adam(adam_learning_rate)
 adam_opt = Adam(adam_learning_rate)
 
 # NatGrads and Adam for VGP with Bernoulli likelihood
 # Stop Adam from optimizing the variational parameters
 set_trainable(vgp_bernoulli_natgrad.q_mu, False)
 set_trainable(vgp_bernoulli_natgrad.q_sqrt, False)
 
 # Create the optimize_tensors for VGP with natural gradients
-natgrad_adam_opt = tf.optimizers.Adam(adam_learning_rate)
 natgrad_adam_opt = Adam(adam_learning_rate)
 natgrad_opt = NaturalGradient(gamma=0.1)
 variational_params = [
     (vgp_bernoulli_natgrad.q_mu, vgp_bernoulli_natgrad.q_sqrt)
@@ -392,7  398,7 @@
 set_trainable(vgp_bernoulli_natgrads_xi.q_sqrt, False)
 
 # Create the optimize_tensors for VGP with Bernoulli likelihood
-adam_opt = tf.optimizers.Adam(adam_learning_rate)
 adam_opt = Adam(adam_learning_rate)
 natgrad_opt = NaturalGradient(gamma=0.01)
 
 variational_params = [

diff --git a/doc/sphinx/notebooks/advanced/variational_fourier_features.pct.py b/doc/sphinx/notebooks/advanced/variational_fourier_features.pct.py
@@ -314,11  314,11 @@ def _conditional_fused(self, Xnew, full_cov, full_output_cov):
         fmean = tf.matmul(A, f, transpose_a=True)
 
         if q_sqrt is not None:
-            if q_sqrt.get_shape().ndims == 2:
             if q_sqrt.shape.ndims == 2:
                 # LTA = A * tf.expand_dims(q_sqrt, 2)  # K x M x N
                 # won't work  # make ticket for this?
                 raise NotImplementedError
-            elif q_sqrt.get_shape().ndims == 3:
             elif q_sqrt.shape.ndims == 3:
                 # L = tf.matrix_band_part(tf.transpose(q_sqrt, (2, 0, 1)), -1, 0)  # K x M x M
 
                 # K x M x N
@@ -331,8  331,7 @@ def _conditional_fused(self, Xnew, full_cov, full_output_cov):
                 ATL = tf.matmul(A, q_sqrt, transpose_a=True)
             else:
                 raise ValueError(
-                    "Bad dimension for q_sqrt: %s"
-                    % str(q_sqrt.get_shape().ndims)
                     "Bad dimension for q_sqrt: %s" % str(q_sqrt.shape.ndims)
                 )
             if full_cov:
                 # fvar = fvar   tf.matmul(LTA, LTA, transpose_a=True)  # K x N x N

diff --git a/gpflow/optimizers/natgrad.py b/gpflow/optimizers/natgrad.py
@@ -201,6  201,8 @@ def __init__(
         """
         name = self.__class__.__name__ if name is None else name
         super().__init__(name)
         # explicitly store name (as TF <2.12 stores it differently from TF 2.12 )
         self.natgrad_name = name
         self.gamma = gamma
         self.xi_transform = xi_transform
 
@@ -263,7  265,7 @@ def _natgrad_steps(
         q_mu_grads, q_sqrt_grads = tape.gradient(loss, [q_mu_vars, q_sqrt_vars])
         # NOTE that these are the gradients in *unconstrained* space
 
-        with tf.name_scope(f"{self._name}/natural_gradient_steps"):
         with tf.name_scope(f"{self.natgrad_name}/natural_gradient_steps"):
             for q_mu_grad, q_sqrt_grad, q_mu, q_sqrt, xi_transform in zip(
                 q_mu_grads, q_sqrt_grads, q_mus, q_sqrts, xis
             ):