/TensorFlow Python



Defined in tensorflow/python/debug/lib/debug_utils.py.

See the guide: TensorFlow Debugger > Functions for adding debug watches

Add debug watches to RunOptions for a TensorFlow graph.

To watch all Tensors on the graph, let both node_name_regex_whitelist and op_type_regex_whitelist be the default (None).

N.B.: 1. Under certain circumstances, the Tensor may not get actually watched (e.g., if the node of the Tensor is constant-folded during runtime). 2. For debugging purposes, the parallel_iteration attribute of all tf.while_loops in the graph are set to 1 to prevent any node from being executed multiple times concurrently. This change does not affect subsequent non-debugged runs of the same tf.while_loops.


  • run_options: An instance of config_pb2.RunOptions to be modified.
  • graph: An instance of ops.Graph.
  • debug_ops: (str or list of str) name(s) of the debug op(s) to use.
  • debug_urls: URLs to send debug values to. Can be a list of strings, a single string, or None. The case of a single string is equivalent to a list consisting of a single string, e.g., file:///tmp/tfdbg_dump_1, grpc://localhost:12345. For debug op types with customizable attributes, each debug op name string can optionally contain a list of attribute names, in the syntax of: debug_op_name(attr_name_1=attr_value_1;attr_name_2=attr_value_2;...)
  • node_name_regex_whitelist: Regular-expression whitelist for node_name, e.g., "(weight_[0-9]+|bias_.*)"
  • op_type_regex_whitelist: Regular-expression whitelist for the op type of nodes, e.g., "(Variable|Add)". If both node_name_regex_whitelist and op_type_regex_whitelist are set, the two filtering operations will occur in a logical AND relation. In other words, a node will be included if and only if it hits both whitelists.
  • tensor_dtype_regex_whitelist: Regular-expression whitelist for Tensor data type, e.g., "^int.*". This whitelist operates in logical AND relations to the two whitelists above.
  • tolerate_debug_op_creation_failures: (bool) whether debug op creation failures (e.g., due to dtype incompatibility) are to be tolerated by not throwing exceptions.
  • global_step: (int) Optional global_step count for this debug tensor watch.

© 2018 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.