flowersteam / Grounding_LLMs_with_online_RL

We perform functional grounding of LLMs' knowledge in BabyAI-Text
MIT License
221 stars 24 forks source link

'NoneType' object has no attribute 'SIGTERM' #1

Closed GoooKuuu closed 1 year ago

GoooKuuu commented 1 year ago

Hello, thank you so much for sharing this amazing work. I am using it for my project and I am encountering an error that I'm hoping you could help me with.

When I run sbatch Symbolic-PPO.slurm, got these error messages:

File "/data/users/xxx/Grounding_LLMs_with_online_RL/experiments/train_symbolic_ppo.py", line 353, in algo = SymbolicPPOAgent(envs, acmodel, args.frames_per_proc, args.discount, args.lr, args.beta1, args.beta2, TypeError: Can't instantiate abstract class SymbolicPPOAgent with abstract method generate_trajectories Exception ignored in: <function ParallelShapedEnv.del at 0x7f7e40689750> Traceback (most recent call last): File "/data/users/xxx/Grounding_LLMs_with_online_RL/babyai-text/babyai/babyai/shaped_env.py", line 188, in del File "/home/xxx/.conda/envs/dlp/lib/python3.10/multiprocessing/process.py", line 133, in terminate File "/home/xxx/.conda/envs/dlp/lib/python3.10/multiprocessing/popen_fork.py", line 57, in terminate AttributeError: 'NoneType' object has no attribute 'SIGTERM'

I would greatly appreciate any help you could provide in resolving this issue. If there's any additional information you need from me, please let me know. Thank you in advance for your time and assistance.

ClementRomac commented 1 year ago

Hi,

My apologies for such a late response! We're glad our work is useful for you.

Concerning your issue, there was a bug in the SymbolicPPOAgent (the generate_trajectories method wasn't implemented). I'm creating a PR to fix this.

ClementRomac commented 1 year ago

Could you please checkout the 1-nonetype-object-has-no-attribute-sigterm and confirm this fixes your issue?

GoooKuuu commented 1 year ago

Thank you for your suggestion:) I checked out the 1-nonetype-object-has-no-attribute-sigterm branch as you recommended, but I still encountered a similar issue with the following error message: Exception ignored in: <function ParallelShapedEnv.del at 0x7f2dd3449000> Traceback (most recent call last): File "/data/users/xxxxx/Grounding_LLMs_with_online_RL/babyai-text/babyai/babyai/shaped_env.py", line 188, in del File "/home/xxxx/dlp/lib/python3.10/multiprocessing/process.py", line 133, in terminate File "/home/xxxxx/dlp/lib/python3.10/multiprocessing/popen_fork.py", line 57, in terminate AttributeError: 'NoneType' object has no attribute 'SIGTERM'

ClementRomac commented 1 year ago

While I'm investigating on this exception, it only occurs when the environment gets closed at the end of training. So that means that you should be able to perform your training anyway.

qizhg commented 5 months ago

Hi @GoooKuuu Did you solve this issue?