TypeError: 'type' object is not subscriptable about comet HOT 6 CLOSED

Looks like I need torch >= 1.4. Somehow poetry install didn't pick up on this, or I have a venv problem. Maybe it should be added to requirements.txt?

from comet.

Hi @mjpost,

I install unbabel-comet in a Python venv 3.7.12. I am also experiencing the same issue:

Traceback (most recent call last):
  File "/home/ubuntu/unbabel-comet-1.0.1/venv/bin/comet-score", line 5, in <module>
    from comet.cli.score import score_command
  File "/home/ubuntu/unbabel-comet-1.0.1/venv/lib/python3.7/site-packages/comet/__init__.py", line 19, in <module>
    from .download_utils import download_model
  File "/home/ubuntu/unbabel-comet-1.0.1/venv/lib/python3.7/site-packages/comet/download_utils.py", line 26, in <module>
    from comet.models import available_metrics
  File "/home/ubuntu/unbabel-comet-1.0.1/venv/lib/python3.7/site-packages/comet/models/__init__.py", line 17, in <module>
    from .regression.regression_metric import RegressionMetric
  File "/home/ubuntu/unbabel-comet-1.0.1/venv/lib/python3.7/site-packages/comet/models/regression/regression_metric.py", line 26, in <module>
    from comet.models.base import CometModel
  File "/home/ubuntu/unbabel-comet-1.0.1/venv/lib/python3.7/site-packages/comet/models/base.py", line 41, in <module>
    class OrderedSampler(Sampler[int]):
TypeError: 'type' object is not subscriptable

when I call comet-score. I installed torch version 1.4.0, however the problem still persists.

Would you know how to fix this?

Thank you for your help.

from comet.

I ended up with torch version 1.10.0. I suggest you double-check that comet-score isn't linking to an old version, maybe installed elsewhere. I think I solved my problems after running pip uninstall both as the regular user and root, and then reinstalling.

from comet.

I just ran into this again today. There were two problems: first, comet 1.0.1 was installed in ~/.local/bin, and comet 0.1.0 in the system bin directory. I had to do pip uninstall unbabel-comet as both my regular user and root. Then, I got the error with pytorch 1.6.0. After upgrading to 1.10.2, all is well.

from comet.

Yes, with torch version 1.10.0, it works very well. Would this mean that the requirements for the package should be adjusted?

from comet.

I would think so. You might try submitting a PR referencing this issue.

from comet.