Hi all,
I’m using cross language programming with Ray Clusters. When I run a sample python file that calls Java on the clusters, with jobconfig specifying the right directory, it works great. But when I try to call the same code from a gymnasium env within ray tune, it fails with the following error. I’m suspecting that it’s just not passing the right jobconfig – I’m setting it in ray.init() before calling tuner.fit(), but the jobconfig doesn’t seem to be passed to the separate training jobs. How do I specify the code_search_path for individual training runs with Ray Tune?
File “/tmp/ray/session_2024-09-04_09-40-15_485206_2286/runtime_resources/working_dir_files/_ray_pkg_0def3cecca026a66/hiv_env.py”, line 79, in init
print(ray.get(calc_class.remote()))
File “/home/ray/anaconda3/lib/python3.9/site-packages/ray/remote_function.py”, line 139, in _remote_proxy
return self._remote(args=args, kwargs=kwargs, **self._default_options)
File “/home/ray/anaconda3/lib/python3.9/site-packages/ray/remote_function.py”, line 416, in _remote
_task_launch_hook(self._function_descriptor, resources, scheduling_strategy)
File “/home/ray/anaconda3/lib/python3.9/site-packages/ray/train/_internal/session.py”, line 547, in _tune_task_and_actor_launch_hook
name = fn.module_name + “.” + fn.class_name + “.” + fn.function_name
AttributeError: ‘ray._raylet.JavaFunctionDescriptor’ object has no attribute ‘module_name’