Skip to content

Fix Atari environment issues (#289) #438

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jacobarrio wants to merge 2 commits into
base: 3.0
Choose a base branch
from
Open

Fix Atari environment issues (#289) #438

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9ef4c7b
fix atari env issues - import path, ale-py version, dynamic workers
jacobarrio Dec 22, 2025
532799a
fix auto worker detection to use CPU cores instead of num_envs
jacobarrio Dec 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion pufferlib/config/atari.ini
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@ rnn_name = Recurrent

[vec]
num_envs = 128
num_workers = 16
num_workers = 0
Copy link
Copy Markdown

@Zentrik Zentrik Dec 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more. 

╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ /puffertank/venv/bin/puffer:10 in <module>                                                       │
│                                                                                                  │
│    7 │   │   sys.argv[0] = sys.argv[0][:-11]                                                     │
│    8 │   elif sys.argv[0].endswith(".exe"):                                                      │
│    9 │   │   sys.argv[0] = sys.argv[0][:-4]                                                      │
│ ❱ 10 │   sys.exit(main())                                                                        │
│   11                                                                                             │
│                                                                                                  │
│ /puffertank/pufferlib/pufferlib/pufferl.py:1332 in main                                          │
│                                                                                                  │
│   1329 │   mode = sys.argv.pop(1)                                                                │
│   1330 │   env_name = sys.argv.pop(1)                                                            │
│   1331 │   if mode == 'train':                                                                   │
│ ❱ 1332 │   │   train(env_name=env_name)                                                          │
│   1333 │   elif mode == 'eval':                                                                  │
│   1334 │   │   eval(env_name=env_name)                                                           │
│   1335 │   elif mode == 'sweep':                                                                 │
│                                                                                                  │
│ /puffertank/pufferlib/pufferlib/pufferl.py:924 in train                                          │
│                                                                                                  │
│    921 │   │   torch.cuda.set_device(local_rank)                                                 │
│    922 │   │   os.environ["CUDA_VISIBLE_DEVICES"] = str(local_rank)                              │
│    923 │                                                                                         │
│ ❱  924 │   vecenv = vecenv or load_env(env_name, args)                                           │
│    925 │   policy = policy or load_policy(args, vecenv, env_name)                                │
│    926 │                                                                                         │
│    927 │   if 'LOCAL_RANK' in os.environ:                                                        │
│                                                                                                  │
│ /puffertank/pufferlib/pufferlib/pufferl.py:1177 in load_env                                      │
│                                                                                                  │
│   1174 │   module_name = 'pufferlib.ocean' if package == 'ocean' else f'pufferlib.environments.  │
│   1175 │   env_module = importlib.import_module(module_name)                                     │
│   1176 │   make_env = env_module.env_creator(env_name)                                           │
│ ❱ 1177 │   return pufferlib.vector.make(make_env, env_kwargs=args['env'], **args['vec'])         │
│   1178                                                                                           │
│   1179 def load_policy(args, vecenv, env_name=''):                                               │
│   1180 │   package = args['package']                                                             │
│                                                                                                  │
│ /puffertank/pufferlib/pufferlib/vector.py:646 in make                                            │
│                                                                                                  │
│   643 │   │   │   kwargs['num_workers'] = num_envs                                               │
│   644 │   │                                                                                      │
│   645 │   │   # TODO: None?                                                                      │
│ ❱ 646 │   │   envs_per_worker = num_envs / kwargs['num_workers']                                 │
│   647 │   │   if envs_per_worker != int(envs_per_worker):                                        │
│   648 │   │   │   raise pufferlib.APIUsageError('num_envs must be divisible by num_workers')     │
│   649                                                                                            │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
ZeroDivisionError: division by zero

Copy link
Copy Markdown

@Zentrik Zentrik Dec 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I changed it to auto but then it defaults to 128 (num_envs) I think which is even worse

Copy link
Copy Markdown
Author

@jacobarrio jacobarrio Dec 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed. Changed it to use auto detection based on CPU cores instead of hardcoding.

Now auto sets num_workers to min(cpu_cores, num_envs) which is safe and performant. Also updated vector.py to handle 'auto' properly instead of defaulting to num_envs.

batch_size = 64

[train]
Expand Down
2 changes: 1 addition & 1 deletion pufferlib/environments/atari/environment.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ def make(name, obs_type='grayscale', frameskip=4,
repeat_action_probability=0.0, render_mode='rgb_array',
buf=None, seed=0):
'''Atari creation function'''
pufferlib.environments.try_import('ale_py', 'AtariEnv')
pufferlib.environments.try_import('ale_py')

ale_render_mode = render_mode
if render_mode == 'human':
Expand Down
2 changes: 1 addition & 1 deletion pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ avalon = [
atari = [
'gymnasium[accept-rom-license]==0.29.1',
'opencv-python==3.4.17.63',
'ale_py==0.9.0',
'ale_py==0.10.1',
]

box2d = [
Expand Down