Closed guillemrbaiges closed 1 year ago
pseudo-rnd-thoughts
commented
1 year ago It looks like
File "/Users/guillem.rossello/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/V11/src/trainer_rl.py", line 50, in __init__
state = self.env.reset()which I suspect will be the return type, not state, info.
As for Poetry, this has been investigated here https://github.com/Farama-Foundation/Gymnasium/pull/86
guillemrbaiges
commented
1 year ago Hi, thanks for the quick reply! Not sure I understand what you mean, I did try both
state = self.env.reset() and state, info = self.env.reset(), although I feel this is not what you mean.
guillemrbaiges
commented
1 year ago I migrated the project to Poetry, this is the dependency tree of the project:
click 8.1.3 Composable command line interface toolkit
└── colorama *
glob2 0.7 Version of the glob module that can capture patterns and supports recursive wildcards
gymnasium 0.28.1 A standard API for reinforcement learning and a diverse set of reference environments (formerly Gym).
├── cloudpickle >=1.2.0
├── farama-notifications >=0.0.1
├── importlib-metadata >=4.8.0
│ └── zipp >=0.5
├── jax-jumpy >=1.0.0
│ └── numpy >=1.18.0
├── numpy >=1.21.0
└── typing-extensions >=4.3.0
matplotlib 3.7.1 Python plotting package
├── contourpy >=1.0.1
│ └── numpy >=1.16
├── cycler >=0.10
├── fonttools >=4.22.0
├── importlib-resources >=3.2.0
│ └── zipp >=3.1.0
├── kiwisolver >=1.0.1
├── numpy >=1.20
├── packaging >=20.0
├── pillow >=6.2.0
├── pyparsing >=2.3.1
└── python-dateutil >=2.7
└── six >=1.5
pygame 2.3.0 Python Game Development
torch 2.0.0 Tensors and Dynamic neural networks in Python with strong GPU acceleration
├── filelock *
├── jinja2 *
│ └── markupsafe >=2.0
├── networkx *
├── nvidia-cublas-cu11 11.10.3.66
│ ├── setuptools *
│ └── wheel *
├── nvidia-cuda-cupti-cu11 11.7.101
│ ├── setuptools *
│ └── wheel *
├── nvidia-cuda-nvrtc-cu11 11.7.99
│ ├── setuptools *
│ └── wheel *
├── nvidia-cuda-runtime-cu11 11.7.99
│ ├── setuptools *
│ └── wheel *
├── nvidia-cudnn-cu11 8.5.0.96
│ ├── setuptools *
│ └── wheel *
├── nvidia-cufft-cu11 10.9.0.58
├── nvidia-curand-cu11 10.2.10.91
│ ├── setuptools *
│ └── wheel *
├── nvidia-cusolver-cu11 11.4.0.1
│ ├── setuptools *
│ └── wheel *
├── nvidia-cusparse-cu11 11.7.4.91
│ ├── setuptools *
│ └── wheel *
├── nvidia-nccl-cu11 2.14.3
├── nvidia-nvtx-cu11 11.7.91
│ ├── setuptools *
│ └── wheel *
├── sympy *
│ └── mpmath >=0.19
├── triton 2.0.0
│ ├── cmake *
│ ├── filelock *
│ ├── lit *
│ └── torch * (circular dependency aborted here)
└── typing-extensions *And I'm now hitting the following seed parameter error (that if I'm not mistaken was addressed in previous versions, not even gymnasium versions but gym versions!).
Traceback (most recent call last):
File "/Users/guillem.rossello/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/V11/src/agent.py", line 62, in <module>
run()
File "/Users/guillem.rossello/Library/Caches/pypoetry/virtualenvs/rl-floorplan-VxTjqtjI-py3.9/lib/python3.9/site-packages/click/core.py", line 1130, in __call__
return self.main(*args, **kwargs)
File "/Users/guillem.rossello/Library/Caches/pypoetry/virtualenvs/rl-floorplan-VxTjqtjI-py3.9/lib/python3.9/site-packages/click/core.py", line 1055, in main
rv = self.invoke(ctx)
File "/Users/guillem.rossello/Library/Caches/pypoetry/virtualenvs/rl-floorplan-VxTjqtjI-py3.9/lib/python3.9/site-packages/click/core.py", line 1404, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "/Users/guillem.rossello/Library/Caches/pypoetry/virtualenvs/rl-floorplan-VxTjqtjI-py3.9/lib/python3.9/site-packages/click/core.py", line 760, in invoke
return __callback(*args, **kwargs)
File "/Users/guillem.rossello/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/V11/src/agent.py", line 58, in run
new_training(config=config, timesteps=timesteps, fname_model=fname_model)
File "/Users/guillem.rossello/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/V11/src/training.py", line 16, in training
trainer = TrainerRL(config)
File "/Users/guillem.rossello/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/V11/src/trainer_rl.py", line 50, in __init__
state = self.env.reset()
File "/Users/guillem.rossello/Library/Caches/pypoetry/virtualenvs/rl-floorplan-VxTjqtjI-py3.9/lib/python3.9/site-packages/gymnasium/core.py", line 462, in reset
obs, info = self.env.reset(seed=seed, options=options)
TypeError: reset() got an unexpected keyword argument 'seed'How can this even happen? I'm printing the gymnasium version right before the reset that triggers the error (the same line 50 in trainer_rl.py of the original error in this issue) and it is 0.28.1.
guillemrbaiges
commented
1 year ago Actually I made an error when I reported this "bug": with the originally explained Conda setup and Gymnasium 0.27.1 I hit this same seed error, not the one I shared at the beginning. Just sharing more data to see if we can find the cause of these issues. Thanks beforehand for the support.
pseudo-rnd-thoughts
commented
1 year ago I suspect that the environment was originally written in gym v0.21, I would use the gym compatibility wrapper in gymnasium https://gymnasium.farama.org/content/gym_compatibility/ Otherwise, I would read the migration guide https://gymnasium.farama.org/content/migration-guide/
Describe the bug
I am hitting this error in all my attempts, after upgrading from gym 0.21 (I had a hard dependency on stables-baselines3) to any of the following versions (installing through Conda):
Code example
System info
Gymnasium installed with Conda, project is executed within this Conda environment. I'm on MacOS Monterey, using Python 3.9.13. Showing logs for my last attempt, but as mentioned earlier I've hit this same error with either gym 0.26.1, gymnasium 0.26.3 and gymnasium 0.27.1
~/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/v11 master *2 !5 ?4 ❯ conda list gym intro2research
packages in environment at /Users/guillem.rossello/opt/anaconda3/envs/intro2research:
#
Name Version Build Channel
gymnasium 0.26.3 py39hecd8cb5_0 gymnasium-notices 0.0.1 py39hecd8cb5_0 ~/Desktop/UNI/Q2/IntroToResearch/RL-FloorPlan/v11 master *2 !5 ?4 ❯ pip show gym intro2research WARNING: Package(s) not found: gym
Additional context
Is this a Conda-related issue? I was using Conda because the project I joined was originally using stable-baselines3, but I really want to migrate to Poetry because honestly I see no advantages using Conda (and a lot of engineering pains). Is Poetry a good fit for Gymnasium?
Checklist