Updated TF installation instructions

- Updated documentation to reflect
  inclusion of GPU support in base `tensorflow`
  python package.
- Fixed `nvidia-driver` package name (and updated
  version number).

Co-authored-by: Daniel Rasmussen <[email protected]>

CHANGES.rst

@@ -36,6 +36,7 @@ Release history
   will still be returned from the function as normal. (`#153`_)
 - A warning will now be raised if activation types are passed to
   ``Converter.swap_activations`` that aren't actually in the model. (`#168`_)
+- Updated TensorFlow installation instruction in documentation. (`#170`_)
 
 **Fixed**
 
@@ -56,6 +57,7 @@ Release history
 .. _#161: https://github.com/nengo/nengo-dl/pull/161
 .. _#164: https://github.com/nengo/nengo-dl/pull/164
 .. _#168: https://github.com/nengo/nengo-dl/pull/168
+.. _#170: https://github.com/nengo/nengo-dl/pull/170
 
 3.2.0 (April 2, 2020)
 ---------------------

docs/installation.rst

@@ -33,13 +33,13 @@ perform a developer installation:
 
 Installing TensorFlow
 ---------------------
-Use ``pip install tensorflow`` to install the minimal version of TensorFlow,
-or ``pip install tensorflow-gpu`` to include GPU support.
+Use ``pip install tensorflow`` to install the latest version of TensorFlow. GPU support
+is included in this package as of version 2.1.0.
 
 In order to use TensorFlow with GPU support you will need to install the appropriate
 Nvidia drivers and CUDA/cuDNN. The precise steps for accomplishing this will depend
 on your system. On Linux the correct Nvidia drivers (as of TensorFlow 2.2.0) can be
-installed via ``sudo apt install nvidia-drivers-430``, and on Windows simply using the
+installed via ``sudo apt install nvidia-driver-440``, and on Windows simply using the
 most up-to-date drivers should work.  For CUDA/cuDNN we recommend using
 `conda <https://docs.conda.io/projects/conda/en/latest/user-guide/install/>`_ to
 simplify the process. ``conda install tensorflow-gpu`` will install TensorFlow as

nengo_dl/builder.py

@@ -181,7 +181,7 @@ class BuildConfig(
         Smoothing parameter for `~nengo.LIF` gradient approximation.
     cpu_only : bool
         True if TensorFlow is only running on the CPU (because that was
-        specified by the user or because ``tensorflow-gpu`` is not installed).
+        specified by the user or because GPU support is not available).
     rng : `~numpy.random.RandomState`
         Seeded random number generator.
     training : ``tf.Tensor`` (bool)

nengo_dl/simulator.py

@@ -377,7 +377,7 @@ class Simulator:  # pylint: disable=too-many-public-methods
         This specifies the computational device on which the simulation will
         run.  The default is ``None``, which means that operations will be assigned
         according to TensorFlow's internal logic (generally speaking, this means that
-        things will be assigned to the GPU if ``tensorflow-gpu`` is installed,
+        things will be assigned to the GPU if GPU support is available,
         otherwise everything will be assigned to the CPU).  The device can be set
         manually by passing the `TensorFlow device specification
         <https://www.tensorflow.org/api_docs/python/tf/Graph#device>`_ to this
@@ -469,8 +469,9 @@ def __init__(
 
         if device is None and not utils.tf_gpu_installed:
             warnings.warn(
-                "No GPU support detected. It is recommended that you "
-                "install tensorflow-gpu (`pip install tensorflow-gpu`)."
+                "No GPU support detected. See "
+                "https://www.nengo.ai/nengo-dl/installation.html#installing-tensorflow "
+                "for instructions on setting up TensorFlow with GPU support."
             )
             logger.info("Running on CPU")
         else:

nengo_dl/tests/test_simulator.py

@@ -669,7 +669,7 @@ def test_model_passing(Simulator, seed):
 @pytest.mark.parametrize("device", ["/cpu:0", "/gpu:0", None])
 def test_devices(Simulator, device, seed, caplog):
     if device == "/gpu:0" and not utils.tf_gpu_installed:
-        pytest.skip("This test requires tensorflow-gpu")
+        pytest.skip("This test requires GPU support")
 
     caplog.set_level(logging.INFO)
 

nengo_dl/transform_builders.py

@@ -44,7 +44,7 @@ def __init__(self, ops, signals, config):
 
         if not self.conv.channels_last and config.cpu_only:
             # TensorFlow doesn't support channels first on CPU, so if
-            # tensorflow-gpu isn't installed we need to force channels_last
+            # GPU support isn't available we need to force channels_last
             # TODO: check if this is supported in future versions
             warnings.warn(
                 "TensorFlow does not support convolution with "