Fix broken create_lerobot_dataset_card

[helper2424]

What this does

During the recording of dataset for HIL-SERL integration I have found the following issue:

Traceback (most recent call last):.93M/4.54M [00:01<00:00, 4.89MB/s]
  File "/Users/helper2424/Documents/lerobot/lerobot/scripts/control_robot.py", line 577, in <module>      | 3/6 [00:03<00:04,  1.38s/it]
    record(robot, **kwargs)███████████| 6/6 [00:03<00:00,  1.88it/s]
  File "/Users/helper2424/Documents/lerobot/lerobot/common/robot_devices/utils.py", line 28, in wrapper
    raise e
  File "/Users/helper2424/Documents/lerobot/lerobot/common/robot_devices/utils.py", line 24, in wrapper
    return func(robot, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/helper2424/Documents/lerobot/lerobot/scripts/control_robot.py", line 334, in record
    dataset.push_to_hub(tags=tags)
  File "/Users/helper2424/Documents/lerobot/lerobot/common/datasets/lerobot_dataset.py", line 499, in push_to_hub
    card = create_lerobot_dataset_card(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/helper2424/Documents/lerobot/lerobot/common/datasets/utils.py", line 500, in create_lerobot_dataset_card
    return DatasetCard.from_template(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/helper2424/Library/Caches/pypoetry/virtualenvs/lerobot-_GqtL9sV-py3.12/lib/python3.12/site-packages/huggingface_hub/repocard.py", line 480, in from_template
    return super().from_template(card_data, template_path, template_str, **template_kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/helper2424/Library/Caches/pypoetry/virtualenvs/lerobot-_GqtL9sV-py3.12/lib/python3.12/site-packages/huggingface_hub/repocard.py", line 329, in from_template
    template_str = Path(template_path).read_text()
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/[email protected]/3.12.7_1/Frameworks/Python.framework/Versions/3.12/lib/python3.12/pathlib.py", line 1027, in read_text
    with self.open(mode='r', encoding=encoding, errors=errors) as f:
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/[email protected]/3.12.7_1/Frameworks/Python.framework/Versions/3.12/lib/python3.12/pathlib.py", line 1013, in open
    return io.open(self, mode, buffering, encoding, errors, newline)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
FileNotFoundError: [Errno 2] No such file or directory: '<contextlib._GeneratorContextManager object at 0x3181632f0>'

This is issue is relevant in the last main.

Examples:

Title	Label
Fixes #[issue]	(🐛 Bug)

How it was tested

How to test that issue is fixed:

• Run data collection script, mine is the following:

python lerobot/scripts/control_robot.py record \
    --robot-path lerobot/configs/robot/koch_helper2424.yaml \
    --num-episodes 2 \
    --warmup-time-s 5 \
    --episode-time-s 10 \
    --fps 30 \
    --assign-rewards 1 \
    --policy-overrides "device=mps" \
    --repo-id ${HF_USER}/hil-serl-pusht-reward-classifier-test \
    --single-task "Push small T object to the correct position"

• Run pytest tests/lerobot/common/datasets/test_utils.py

[jackvial]

Tested on my setup and it worked great. Here's the tests dataset I created https://huggingface.co/datasets/jackvial/create_lerobot_dataset_card_fix_test_0

The fix and unit tests make sense 👍

[michel-aractingi]

cc @aliberts for visibility

[aliberts]

Thank you @helper2424!

Could you expand on why using

card_template = (importlib.resources.files("lerobot.common.datasets") / "card_template.md").read_text()

DatasetCard.from_template(
    card_data=card_data,
    template_str=card_template,
)

rather than

card_template_path = importlib.resources.path("lerobot.common.datasets", "card_template.md")

DatasetCard.from_template(
    card_data=card_data,
    template_path=str(card_template_path),
)

fixed your issue? Both should be +/- equivalent.
It looked from your stack trace that the file wasn't there. Am I reading this wrong?

[helper2424]

@aliberts Hello, the file exists, the issue with the following - code str(card_template_path) returns '<contextlib._GeneratorContextManager object at 0x3181632f0>'. It return the introspection of the object instead of the path to the file. Also, I see that the contextlib generates a lot of warnings about the API usage, so I decided the way that doesn't generate warnings and that looks ok with their long term development way.

[helper2424]

@aliberts in my PR I read the content of the file and pass it to the from_template.

[aliberts]

@helper2424 I can't seem to be able to reproduce you error. Currently, running this on main give me no error:

from lerobot.common.datasets.utils import create_lerobot_dataset_card

card = create_lerobot_dataset_card()
print(card)

And when running it with the debugger and inspecting str(card_template_path), I do get the correct path as a str.

After digging a bit, I think this is probably due to differences between our environments and how lerobot was installed, which could make importlib.resources behave differently. I also noticed that importlib.resources.path should be used as a context manager, which it's not here. Let's make this change then!

[aliberts]

Style test is failing, you can fix this by running our pre-commits.
Install pre-commit:

# using pip
pip install -e '.[dev]'

# or using poetry
poetry install --extras 'dev'

Run pre-commits

pre-commit run -a

Then commit & push

[aliberts]

LGTM, thanks!