Numpy version

[duckduckdoof]

Added numpy version lock -- installing AWE_Workbench causes an installation issue when pulling model data. Also updates lexica dependency.

[bradley-erickson]

I'm reluctant to add version locking here without knowing where the code breaks. Have we figured out why we need both this and the torch version lock?
Ideally, we don't have to lock either and we just fix the code that is breaking.

[bradley-erickson]

• Mark TODOs in the code where the version locks are needed (like our chosen model, wherever its accessed, needs numpy v1)

[duckduckdoof]

I was able to track down the errors which originally caused me to revert the numpy version, but it's more confusing that I thought:

File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/awe_languagetool/setup/data.py", line 29, in <module>
    from spacy.cli.download import download
  File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/spacy/__init__.py", line 6, in <module>
    from .errors import setup_default_warnings
  File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/spacy/errors.py", line 2, in <module>
    from .compat import Literal
  File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/spacy/compat.py", line 38, in <module>
    from thinc.api import Optimizer  # noqa: F401
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/thinc/api.py", line 1, in <module>
    from .backends import (
  File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/thinc/backends/__init__.py", line 17, in <module>
    from .cupy_ops import CupyOps
  File "/home/goose/anaconda3/envs/arglab-wb-test/lib/python3.11/site-packages/thinc/backends/cupy_ops.py", line 16, in <module>
    from .numpy_ops import NumpyOps
  File "thinc/backends/numpy_ops.pyx", line 1, in init thinc.backends.numpy_ops
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

Components is what installs numpy, but the issue is invoked after trying the python -m awe_workbench.setup.data --develop command

[duckduckdoof]

I'm not sure - referencing #6 (comment) - if there is a specific place in Components I can comment the proper TODO?

[bradley-erickson]

The version lock is 1.26.4. Is this just a typo?

[bradley-erickson]

Is this the correct spot for this comment? I thought it had to do with the model we are using. So this comment ought to be placed where the model is loaded in. A quick search in the codebase makes me think it belongs in awe_components/wordprobs/wordProbabilityInContext.py; however, this may not be the correct spot for it.

[duckduckdoof]

The version lock is 1.26.4. Is this just a typo?

Yup! I'll fix that.

Is this the correct spot for this comment? I thought it had to do with the model we are using. So this comment ought to be placed where the model is loaded in. A quick search in the codebase makes me think it belongs in awe_components/wordprobs/wordProbabilityInContext.py; however, this may not be the correct spot for it.

From what I remember about the errors I was getting, it was referencing the spacy.cli.download import in data.py. I'll look at awe_components/wordprobs/wordProbabilityInContext.py nonetheless, but I think my comment is in the right place.