[TF] Change eager mode (#1763)

doctr/file_utils.py

@@ -80,10 +80,7 @@ def ensure_keras_v2() -> None:  # pragma: no cover
         else:
             logging.info(f"TensorFlow version {_tf_version} available.")
             ensure_keras_v2()
-            import tensorflow as tf
 
-            # Enable eager execution - this is required for some models to work properly
-            tf.config.run_functions_eagerly(True)
 else:  # pragma: no cover
     logging.info("Disabling Tensorflow because USE_TORCH is set")
     _tf_available = False

doctr/models/recognition/parseq/tensorflow.py

@@ -250,7 +250,6 @@ def decode(
         target_query = self.dropout(target_query, **kwargs)
         return self.decoder(target_query, content, memory, target_mask, **kwargs)
 
-    @tf.function
     def decode_autoregressive(self, features: tf.Tensor, max_len: Optional[int] = None, **kwargs) -> tf.Tensor:
         """Generate predictions for the given features."""
         max_length = max_len if max_len is not None else self.max_length

doctr/models/utils/tensorflow.py

@@ -161,6 +161,10 @@ def export_model_to_onnx(
     -------
         the path to the exported model and a list with the output layer names
     """
+    # get the users eager mode
+    eager_mode = tf.executing_eagerly()
+    # set eager mode to true to avoid issues with tf2onnx
+    tf.config.run_functions_eagerly(True)
     large_model = kwargs.get("large_model", False)
     model_proto, _ = tf2onnx.convert.from_keras(
         model,
@@ -171,6 +175,9 @@ def export_model_to_onnx(
     # Get the output layer names
     output = [n.name for n in model_proto.graph.output]
 
+    # reset the eager mode to the users mode
+    tf.config.run_functions_eagerly(eager_mode)
+
     # models which are too large (weights > 2GB while converting to ONNX) needs to be handled
     # about an external tensor storage where the graph and weights are seperatly stored in a archive
     if large_model: