Reload weights after plateau

.gitignore

@@ -37,3 +37,4 @@ Dockerfile.train
 doc/xml-c
 doc/xml-java
 doc/xml-dotnet
+convert_graphdef_memmapped_format

training/mozilla_voice_stt_training/train.py

@@ -29,7 +29,7 @@
 from .evaluate import evaluate
 from six.moves import zip, range
 from .util.config import Config, initialize_globals
-from .util.checkpoints import load_or_init_graph_for_training, load_graph_for_evaluation
+from .util.checkpoints import load_or_init_graph_for_training, load_graph_for_evaluation, reload_best_checkpoint
 from .util.evaluate_tools import save_samples_json
 from .util.feeding import create_dataset, audio_to_features, audiofile_to_features
 from .util.flags import create_flags, FLAGS
@@ -645,6 +645,9 @@ def __call__(self, progress, data, **kwargs):
                         log_info('Encountered a plateau, reducing learning rate to {}'.format(
                             current_learning_rate))
 
+                        # Reload checkpoint that we use the best_dev weights again
+                        reload_best_checkpoint(session)
+
                 if FLAGS.metrics_files:
                     # Read only metrics, not affecting best validation loss tracking
                     for source, init_op in zip(metrics_sources, metrics_init_ops):

training/mozilla_voice_stt_training/util/checkpoints.py

@@ -118,6 +118,10 @@ def _load_or_init_impl(session, method_order, allow_drop_layers):
     sys.exit(1)
 
 
+def reload_best_checkpoint(session):
+    _load_or_init_impl(session, ['best'], allow_drop_layers=False)
+
+
 def load_or_init_graph_for_training(session):
     '''
     Load variables from checkpoint or initialize variables. By default this will