Disable grappler remapping, to avoid introducing unsupported ops. Also: avoid removing the input nodes of a graph.

PiperOrigin-RevId: 226712843

tensorflow/lite/python/lite.py

@@ -71,11 +71,12 @@ from tensorflow.python.util import deprecation as _deprecation
 from tensorflow.python.util.tf_export import tf_export as _tf_export
 
 
-def _run_graph_optimizations(graph_def, output_arrays):
+def _run_graph_optimizations(graph_def, input_arrays, output_arrays):
   """Apply standard TensorFlow optimizations to the graph_def.
 
   Args:
     graph_def: Frozen GraphDef to be optimized.
+    input_arrays: List of arrays that are considered inputs of the graph.
     output_arrays: List of arrays that are considered outputs of the graph.
 
   Returns:
@@ -86,13 +87,16 @@ def _run_graph_optimizations(graph_def, output_arrays):
   # We need to add a collection called 'train_op' so that grappler
   # knows what the outputs are.
   fetch_collection = _meta_graph_pb2.CollectionDef()
-  for output in output_arrays:
-    fetch_collection.node_list.value.append(output)
+  for array in input_arrays + output_arrays:
+    fetch_collection.node_list.value.append(array.name)
   meta_graph.collection_def["train_op"].CopyFrom(fetch_collection)
 
   config = _config_pb2.ConfigProto()
   rewrite_options = config.graph_options.rewrite_options
   rewrite_options.layout_optimizer = _rewriter_config_pb2.RewriterConfig.ON
+  # Avoid remapping as it creates ops like _FusedConv2D, which are not
+  # supported by TF Lite.
+  rewrite_options.remapping = _rewriter_config_pb2.RewriterConfig.OFF
   return _tf_optimizer.OptimizeGraph(config, meta_graph)
 
 
@@ -482,7 +486,7 @@ class TFLiteConverter(object):
     else:
       try:
         optimized_graph = _run_graph_optimizations(
-            self._graph_def, [t.name for t in self._output_tensors])
+            self._graph_def, self._input_tensors, self._output_tensors)
       except Exception:
         optimized_graph = self._graph_def