AttributeError: 'NoneType' object has no attribute 'is_initialized'

RLlib
1

Hi all,
Has anyone encountered such an error for no reason?

I am pretty sure there is no issue in the code, as it was working pretty well for the 1st trial including 200 iters; and in the second trial it worked till 175 iters, but then suddenly gave this error. I am running exactly the same code on the Linux machine without any errors. But when I run on Windows I get this error.

Thanks!

2023-07-17 00:52:45,667 WARNING tune_controller.py:271 -- Trial controller checkpointing failed: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\\Users\\team\\ray_results\\PPO\\.tmp_generator' -> 'C:\\Users\\team\\ray_results\\PPO\\basic-variant-state-2023-07-16_22-24-44.json'
Traceback (most recent call last):
  File "D:\Reza\housing_design\rlb_agents\wb_runner.py", line 145, in <module>
    self.run()
  File "D:\Reza\housing_design\rlb_agents\wb_runner.py", line 129, in run
    self.learner.learn(trial_id, run_id, execution)
  File "D:\Reza\housing_design\rlb_agents\learner.py", line 106, in learn
    self.tunner.tune() # self.checkpoint_dir = self.tunner.tune()
  File "D:\Reza\housing_design\rlb_agents\tunner.py", line 100, in tune
    results = tuner.fit()
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\tuner.py", line 347, in fit
    return self._local_tuner.fit()
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\impl\tuner_internal.py", line 588, in fit
    analysis = self._fit_internal(trainable, param_space)
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\impl\tuner_internal.py", line 712, in _fit_internal
    analysis = run(
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\tune.py", line 1070, in run
    runner.step()
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\execution\tune_controller.py", line 272, in step
    raise e
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\execution\tune_controller.py", line 269, in step
    self.checkpoint()
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\execution\trial_runner.py", line 491, in checkpoint
    self._checkpoint_manager.checkpoint(
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\execution\experiment_state.py", line 232, in checkpoint
    save_fn()
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\execution\trial_runner.py", line 383, in save_to_dir
    self._search_alg.save_to_dir(
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\search\basic_variant.py", line 409, in save_to_dir
    _atomic_save(
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\tune\utils\util.py", line 515, in _atomic_save
    os.replace(tmp_search_ckpt_path, os.path.join(checkpoint_dir, file_name))
PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\\Users\\team\\ray_results\\PPO\\.tmp_generator' -> 'C:\\Users\\team\\ray_results\\PPO\\basic-variant-state-2023-07-16_22-24-44.json'
Exception ignored in: <function WandbLoggerCallback.__del__ at 0x0000027D036D9AB0>
Traceback (most recent call last):
  File "C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\air\integrations\wandb.py", line 727, in __del__
AttributeError: 'NoneType' object has no attribute 'is_initialized'
2

Hello,
I am running into this same problem at all runs and am not able to fix it. Were you able to solve it?

3

Hi @jaroan ,
sorry, but I don’t remember how I fixed the issue at that time :frowning:

4

To my understanding, the title and the topic of this thread is misleading a bit.

The following part of the stack trace

Exception ignored in: <function WandbLoggerCallback.del at 0x0000027D036D9AB0>
Traceback (most recent call last):
File “C:\Users\team\miniconda3\envs\ray\lib\site-packages\ray\air\integrations\wandb.py”, line 727, in del
AttributeError: ‘NoneType’ object has no attribute ‘is_initialized’

is coming quite frequently and I observe it in wandb 0.18.6 and 0.18.7 used with ray 2.10.

The PermissionError beforehand is rather the point to worry about.