-
Notifications
You must be signed in to change notification settings - Fork 3
Description
I Followed the setup steps mentioned in the README.
When I run the replay job I get the following error:
Traceback (most recent call last):
File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 1671, in
main()
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 1661, in main
benchmark_result = replay_manager.benchTime()
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 1301, in benchTime
self.preprocess_graph()
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 1279, in preprocess_graph
self.extract_subgraph(root)
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 485, in extract_subgraph
success, msg = analayze_node(node)
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 451, in analayze_node
func, output_count = self.build_func(node)
File "/home/amir/chakra_replay/et_replay/tools/et_replay.py", line 774, in build_func
func, output_count = build_torchscript_func(node)
File "/home/amir/chakra_replay/et_replay/et_replay_utils.py", line 154, in build_torchscript_func
r"^.*?(", "", input_types[0]
IndexError: list index out of range
Adding the following print to et_replay_utils.py:
print (n.op_schema)
# items = [item for item in tmp[0].split(',') if item != ' *']
types = [item for item in tmp[0].split(" ") if "," not in item][:-1]
print(n.name, n.id, types)
I get the following output:
mkldnn::_is_mkldnn_acl_supported() -> bool _0
mkldnn::_is_mkldnn_acl_supported 75704 []
Looks like the code doesn't handle operators without input arguments.