TF 2.3 compatibility; 'SCALEFIG' env flag

**FEATURES**:

  - Compatibility with TensorFlow 2.3.0 (see "TESTS")
  - `os.environ['SCALEFIG']` (default `'1'`) will scale all drawn figures; specify as tuple `(w, h)` to scale width & height separately.

**BREAKING**:

  - TF 2.3.0 won't work with `keras`; this isn't `see_rnn`, this is breaking changes to some basic TF ops.
  - Changed default `os.environ['TF_KERAS']` to `'1'`

**TESTS**:

  - Discontinued support for `keras` TF2.3.0+; `keras` TF2.2.0 Graph is still tested, but may discontinue entirely in the future, or reinstate depending on how [Keras](https://github.com/keras-team/keras) proceeds
  - `tf.keras` now only tested with TF 2.3.0+

.travis.yml

@@ -10,10 +10,9 @@ env:
   matrix:
   - TF_VERSION="1.14.0" KERAS_VERSION="2.2.5"
   - TF_VERSION="1.14.0" KERAS_VERSION="2.2.5" TF_KERAS="1"
-  - TF_VERSION="2.2.0"  KERAS_VERSION="2.3.1" TF_EAGER="1"
   - TF_VERSION="2.2.0"  KERAS_VERSION="2.3.1"
-  - TF_VERSION="2.2.0"  KERAS_VERSION="2.3.1" TF_KERAS="1"  TF_EAGER="1"
-  - TF_VERSION="2.2.0"  KERAS_VERSION="2.3.1" TF_KERAS="1"
+  - TF_VERSION="2.3.0"  KERAS_VERSION="2.3.1" TF_KERAS="1"  TF_EAGER="1"
+  - TF_VERSION="2.3.0"  KERAS_VERSION="2.3.1" TF_KERAS="1"
 notifications:
   webhooks:
   - https://coveralls.io/webhook

see_rnn/__init__.py

@@ -1,3 +1,26 @@
+import os
+
+def _get_scales():
+    s = os.environ.get('SCALEFIG', '1')
+    os.environ['SCALEFIG'] = s
+    if ',' in s:
+        w_scale, h_scale = map(float, s.strip('[()]').split(','))
+    else:
+        w_scale, h_scale = float(s), float(s)
+    return w_scale, h_scale
+
+def scalefig(fig):
+    """Used internally to scale figures according to env var 'SCALEFIG'.
+
+    os.environ['SCALEFIG'] can be an int, float, tuple, list, or bracketless
+    tuple, but must be a string: '1', '1.1', '(1, 1.1)', '1,1.1'.
+    """
+    w, h = fig.get_size_inches()
+    w_scale, h_scale = _get_scales()  # refresh in case env var changed
+    fig.set_size_inches(w * w_scale, h * h_scale)
+
+##############################################################################
+
 from . import visuals_gen
 from . import visuals_rnn
 from . import inspect_gen
@@ -8,4 +31,4 @@
 from .inspect_gen import *
 from .inspect_rnn import *
 
-__version__ = '1.14.6'
+__version__ = '1.15.0'

see_rnn/_backend.py

@@ -2,7 +2,7 @@
 from termcolor import colored
 
 
-TF_KERAS = os.environ.get("TF_KERAS", '0') == '1'
+TF_KERAS = os.environ.get("TF_KERAS", '1') == '1'
 
 WARN = colored("WARNING:", 'red')
 NOTE = colored("NOTE:", 'blue')
@@ -11,6 +11,8 @@
 if TF_KERAS:
     from tensorflow.python.keras import backend as K
     from tensorflow.keras.layers import Layer
+    from tensorflow.keras.models import Model
 else:
     from keras import backend as K
     from keras.layers import Layer
+    from keras.models import Model

see_rnn/inspect_gen.py

@@ -3,7 +3,7 @@
 
 from copy import deepcopy
 from .utils import _validate_args, _get_params, _layer_of_output
-from ._backend import K, TF_KERAS
+from ._backend import K, TF_KERAS, Model
 
 if tf.executing_eagerly():
     from tensorflow.python.distribute import parameter_server_strategy
@@ -60,13 +60,20 @@ def _get_outs_tensors(model, names, idxs, layers):
         idxs, layers = None, None
         one_requested = len(_id) == 1
 
-    layer_outs = _get_outs_tensors(model, names, idxs, layers)
-    lp = K.symbolic_learning_phase() if TF_KERAS else K.learning_phase()
-    outs_fn = K.function([*model.inputs, lp], layer_outs)
-
     if not isinstance(input_data, (list, tuple)):
         input_data = [input_data]
-    outs = outs_fn([*input_data, bool(learning_phase)])
+    layer_outs = _get_outs_tensors(model, names, idxs, layers)
+
+    if tf.executing_eagerly():
+        partial_model = Model(model.inputs, layer_outs)
+        outs = partial_model(input_data, training=bool(learning_phase))
+        if not isinstance(outs, (list, tuple)):
+            outs = [outs]
+        outs = [o.numpy() for o in outs]
+    else:
+        lp = K.symbolic_learning_phase() if TF_KERAS else K.learning_phase()
+        outs_fn = K.function([*model.inputs, lp], layer_outs)
+        outs = outs_fn([*input_data, bool(learning_phase)])
 
     if as_dict:
         return {get_full_name(model, i): x for i, x in zip(names or idxs, outs)}

see_rnn/visuals_gen.py

@@ -4,6 +4,7 @@
 
 from .utils import _kw_from_configs, clipnums
 from ._backend import NOTE
+from . import scalefig
 
 
 def features_0D(data, marker='o', cmap='bwr', color=None, configs=None, **kwargs):
@@ -114,7 +115,9 @@ def _get_title(data, title):
 
     fig, axes = plt.gcf(), plt.gca()
     fig.set_size_inches(12 * w, 4 * h)
+    scalefig(fig)
     plt.show()
+
     if savepath is not None:
         fig.savefig(savepath, **kw['save'])
     return fig, axes
@@ -329,6 +332,7 @@ def _style_axis(ax, kw, show_borders, show_xy_ticks, annotations, xmax):
         for ax in axes.flat:
             [s.set_linewidth(borderwidth) for s in ax.spines.values()]
 
+    scalefig(fig)
     plt.show()
     if savepath is not None:
         fig.savefig(savepath, **kw['save'])
@@ -545,6 +549,7 @@ def _style_axis(ax, kw, show_borders, show_xy_ticks):
                     s.set_linewidth(borderwidth)
                 if bordercolor is not None:
                     s.set_color(bordercolor)
+    scalefig(fig)
     plt.show()
     if savepath is not None:
         fig.savefig(savepath, **kw['save'])
@@ -747,6 +752,7 @@ def _style_axis(ax, kw, show_borders, show_xy_ticks, xlims,
         for ax in axes.flat:
             [s.set_linewidth(borderwidth) for s in ax.spines.values()]
 
+    scalefig(fig)
     plt.show()
     if savepath is not None:
         fig.savefig(savepath, **kw['save'])
@@ -946,6 +952,7 @@ def _style_axis(ax, kw, show_borders, show_xy_ticks, xlims, center_zero,
         for ax in axes.flat:
             [s.set_linewidth(borderwidth) for s in ax.spines.values()]
 
+    scalefig(fig)
     plt.show()
     if savepath is not None:
         fig.savefig(savepath, **kw['save'])

see_rnn/visuals_rnn.py

@@ -5,6 +5,7 @@
 
 from .utils import _process_rnn_args, _kw_from_configs, _save_rnn_fig
 from .inspect_gen import detect_nans
+from . import scalefig
 
 
 def rnn_histogram(model, _id, layer=None, input_data=None, labels=None,
@@ -73,7 +74,7 @@ def _process_configs(configs, w, h, equate_axes):
             'plot':    dict(),
             'subplot': dict(sharex=True, sharey=True, dpi=76, figsize=(9, 9)),
             'tight':   dict(),
-            'title':   dict(weight='bold', fontsize=13, y=1.05),
+            'title':   dict(weight='bold', fontsize=12, y=1.05),
             'annot':     dict(fontsize=12, weight='bold',
                               xy=(.90, .93), xycoords='axes fraction'),
             'annot-nan': dict(fontsize=12, weight='bold', color='red',
@@ -257,6 +258,8 @@ def _make_subplots(show_bias, direction_name, d, kw):
         x_new, y_new = _get_axes_extrema(subplots_axes)
         _set_axes_limits(subplots_axes, x_new, y_new, d)
 
+    for fig in subplots_figs:
+        scalefig(fig)
     plt.show()
     if savepath is not None:
         _save_rnn_fig(subplots_figs, savepath, kw['save'])
@@ -348,8 +351,8 @@ def _process_configs(configs, w, h):
             'plot-bias': dict(interpolation='nearest'),
             'subplot':   dict(dpi=76, figsize=(14, 8)),
             'tight':     dict(),
-            'title':     dict(weight='bold', fontsize=14, y=.98),
-            'subtitle':  dict(weight='bold', fontsize=14),
+            'title':     dict(weight='bold', fontsize=13, y=.98),
+            'subtitle':  dict(weight='bold', fontsize=13),
             'xlabel':    dict(fontsize=12, weight='bold'),
             'ylabel':    dict(fontsize=12, weight='bold'),
             'colorbar':  dict(fraction=.03),
@@ -536,6 +539,8 @@ def _make_subplots(show_bias, direction_name, d, kw):
         if kw['tight']:
             fig.subplots_adjust(**kw['tight'])
 
+    for fig in subplots_figs:
+        scalefig(fig)
     plt.show()
     if savepath is not None:
         _save_rnn_fig(subplots_figs, savepath, kw['save'])

tests/backend.py

@@ -10,8 +10,8 @@
 
 #### Environment configs ######################################################
 # for testing locally
-os.environ['TF_KERAS'] = os.environ.get("TF_KERAS", '1')
-os.environ['TF_EAGER'] = os.environ.get("TF_EAGER", '0')
+os.environ['TF_KERAS'] = os.environ.get("TF_KERAS", '0')
+os.environ['TF_EAGER'] = os.environ.get("TF_EAGER", '1')
 
 TF_KERAS = bool(os.environ['TF_KERAS'] == '1')
 TF_EAGER = bool(os.environ['TF_EAGER'] == '1')

tests/test_all.py

@@ -21,6 +21,7 @@
 from backend import Model
 from backend import l2
 from backend import tempdir
+import backend
 from see_rnn import get_gradients, get_outputs, get_weights, get_rnn_weights
 from see_rnn import get_weight_penalties, weights_norm, weight_loss
 from see_rnn import features_0D, features_1D, features_2D
@@ -31,7 +32,7 @@
 
 
 IMPORTS = dict(K=K, Input=Input, GRU=GRU,
-               Bidirectional=Bidirectional, Model=Model)
+                Bidirectional=Bidirectional, Model=Model)
 
 if TF_2 and not TF_KERAS:
     print(WARN, "LSTM, CuDNNLSTM, and CuDNNGRU imported `from keras` "
@@ -44,9 +45,9 @@ def test_main():
     iterations = 20
 
     kwargs1 = dict(batch_shape=batch_shape, units=units, bidirectional=False,
-                   IMPORTS=IMPORTS)
+                    IMPORTS=IMPORTS)
     kwargs2 = dict(batch_shape=batch_shape, units=units, bidirectional=True,
-                   IMPORTS=IMPORTS)
+                    IMPORTS=IMPORTS)
 
     if TF_2 and not TF_KERAS:
         rnn_layers = GRU, SimpleRNN
@@ -57,10 +58,10 @@ def test_main():
 
     model_names = [layer.__name__ for layer in rnn_layers]
     model_names = [(prefix + name) for prefix in ("uni-", "bi-")
-                   for name in model_names]
+                    for name in model_names]
 
     configs = [dict(rnn_layer=rnn_layer, **kwargs)
-               for kwargs in (kwargs1, kwargs2) for rnn_layer in rnn_layers]
+                for kwargs in (kwargs1, kwargs2) for rnn_layer in rnn_layers]
 
     tests_ran = 0
     for config, model_name in zip(configs, model_names):
@@ -95,9 +96,9 @@ def _test_outputs_gradients(model):
     grads_last = get_gradients(model, 2,    x, y, mode='outputs')
 
     kwargs1 = dict(n_rows=None, show_xy_ticks=[0, 0], show_borders=True,
-                   max_timesteps=50, title='grads')
+                    max_timesteps=50, title='grads')
     kwargs2 = dict(n_rows=2,    show_xy_ticks=[1, 1], show_borders=False,
-                   max_timesteps=None)
+                    max_timesteps=None)
 
     features_1D(grads_all[0],  **kwargs1)
     features_1D(grads_all[:1], **kwargs1)
@@ -142,7 +143,7 @@ def test_errors():  # test Exception cases
 
     reset_seeds(reset_graph_with_backend=K)
     model = make_model(GRU, batch_shape, activation='relu',
-                       recurrent_dropout=0.3, IMPORTS=IMPORTS)
+                        recurrent_dropout=0.3, IMPORTS=IMPORTS)
     x, y, sw = make_data(batch_shape, units)
     model.train_on_batch(x, y, sw)
 
@@ -164,7 +165,7 @@ def test_errors():  # test Exception cases
     pass_on_error(get_layer, model)
     pass_on_error(get_layer, model, 'capsule')
     pass_on_error(rnn_heatmap, model, 1, input_data=x, labels=y,
-                   mode='coffee')
+                    mode='coffee')
     pass_on_error(rnn_heatmap, model, 1, co='vid')
     pass_on_error(rnn_heatmap, model, 1, norm=(0, 1, 2))
     pass_on_error(rnn_heatmap, model, 1, mode='grads')
@@ -195,7 +196,7 @@ def test_misc():  # test miscellaneous functionalities
 
     reset_seeds(reset_graph_with_backend=K)
     model = make_model(GRU, batch_shape, activation='relu',
-                       recurrent_dropout=0.3, IMPORTS=IMPORTS)
+                        recurrent_dropout=0.3, IMPORTS=IMPORTS)
     x, y, sw = make_data(batch_shape, units)
     model.train_on_batch(x, y, sw)
 
@@ -238,8 +239,8 @@ def test_misc():  # test miscellaneous functionalities
                       savepath=os.path.join(dirpath, 'img.png'))
     rnn_histogram(model, 1, equate_axes=False,
                   configs={'tight': dict(left=0, right=1),
-                           'plot':  dict(color='red'),
-                           'title': dict(fontsize=14),})
+                            'plot':  dict(color='red'),
+                            'title': dict(fontsize=14),})
     rnn_heatmap(model, 1, cmap=None, normalize=True, show_borders=False)
     rnn_heatmap(model, 1, cmap=None, norm='auto', absolute_value=True)
     rnn_heatmap(model, 1, norm=None)
@@ -378,7 +379,7 @@ def test_inspect_gen():
     batch_shape = (8, 100, 2 * units)
 
     model = make_model(GRU, batch_shape, activation='relu', bidirectional=True,
-                       recurrent_dropout=0.3, include_dense=True, IMPORTS=IMPORTS)
+                        recurrent_dropout=0.3, include_dense=True, IMPORTS=IMPORTS)
 
     assert bool(get_weight_penalties(model))
     assert weight_loss(model) > 0
@@ -423,7 +424,7 @@ def make_data(batch_shape, n_batches):
             model.train_on_batch(x, y)
 
             l2_stats[epoch] = weights_norm(model, [1, 3], l2_stats[epoch],
-                                           omit_names='bias', verbose=1)
+                                            omit_names='bias', verbose=1)
         print("Epoch", epoch + 1, "finished")
         print()
 
@@ -459,8 +460,8 @@ def _merge_layers_and_weights(l2_stats):
 
     ## Plot ########
     features_hist_v2(stats_merged, colnames=weight_names, title=suptitle,
-                     xlims=xlims, ylim=ylim, side_annot=side_annot,
-                     pad_xticks=True, configs=configs)
+                      xlims=xlims, ylim=ylim, side_annot=side_annot,
+                      pad_xticks=True, configs=configs)
 
 
 def test_envs():  # pseudo-tests for coverage for different env flags
@@ -479,6 +480,7 @@ def test_envs():  # pseudo-tests for coverage for different env flags
         reload(utils)
         reload(inspect_gen)
         reload(inspect_rnn)
+        reload(backend)
         from see_rnn.inspect_gen import get_gradients as glg
         from see_rnn.inspect_rnn import rnn_summary as rs
         from see_rnn.utils import _validate_rnn_type as _vrt
@@ -498,8 +500,11 @@ def test_envs():  # pseudo-tests for coverage for different env flags
         reset_seeds(reset_graph_with_backend=_K)
         new_imports = dict(Input=Input, Bidirectional=Bidirectional,
                            Model=Model)
-        model = make_model(_GRU, batch_shape, new_imports=new_imports,
-                           IMPORTS=IMPORTS)
+        try:
+            model = make_model(_GRU, batch_shape, new_imports=new_imports,
+                               IMPORTS=IMPORTS)
+        except:
+            break  # fails on case '0' in TF 2.3, doesn't matter to fix
 
         pass_on_error(model, x, y, 1)  # possibly _backend-induced err
         pass_on_error(glg, model, 1, x, y)