error freezing model for se_ar

[negf]

Hi,

Trying to freeze a model trained with se_ar throws the following error
Use tf.compat.v1.graph_util.extract_sub_graph
Traceback (most recent call last):
File "/home/deepmd-kit/bin/dp", line 10, in
sys.exit(main())
File "/home/deepmd-kit/lib/python3.7/site-packages/deepmd/main.py", line 65, in main
freeze(args)
File "/home/deepmd-kit/lib/python3.7/site-packages/deepmd/freeze.py", line 99, in freeze
freeze_graph(args.folder, args.output, args.nodes)
File "/home/deepmd-kit/lib/python3.7/site-packages/deepmd/freeze.py", line 89, in freeze_graph
output_node_names.split(",") # The output node names are used to select the usefull nodes
File "/home/deepmd-kit/lib/python3.7/site-packages/tensorflow_core/python/util/deprecation.py", line 324, in new_func
return func(*args, **kwargs)
File "/home/deepmd-kit/lib/python3.7/site-packages/tensorflow_core/python/framework/graph_util_impl.py", line 275, in convert_variables_to_constants
inference_graph = extract_sub_graph(input_graph_def, output_node_names)
File "/home/deepmd-kit/lib/python3.7/site-packages/tensorflow_core/python/util/deprecation.py", line 324, in new_func
return func(*args, **kwargs)
File "/home/deepmd-kit/lib/python3.7/site-packages/tensorflow_core/python/framework/graph_util_impl.py", line 195, in extract_sub_graph
_assert_nodes_are_present(name_to_node, dest_nodes)
File "/home/deepmd-kit/lib/python3.7/site-packages/tensorflow_core/python/framework/graph_util_impl.py", line 150, in _assert_nodes_are_present
assert d in name_to_node, "%s is not in graph" % d
AssertionError: descrpt_attr/rcut is not in graph

Any ideas?

Best
M

[amcadmus]

We have fixed this bug on the devel branch, please see the PR #133

[negf]

thanks for the swift reply. i will give it a try.