Talk to Your Documents with Retrieval-Augmented Generation not working

[Velcin]

Hi,

I've tested the "talk to your document feature" via the notebook demo (OnPrem.LLM-QuickStart.ipynb) but it doesn't work. It seems to be during the data "ingestion".

Here is the error I get:

---

AttributeError Traceback (most recent call last) in <cell line: 1>() ----> 1 llm.ingest('/tmp/sample_data')

14 frames /usr/local/lib/python3.10/dist-packages/numpy/testing/_private/utils.py in 55 IS_PYSTON = hasattr(sys, "pyston_version_info") 56 HAS_REFCOUNT = getattr(sys, 'getrefcount', None) is not None and not IS_PYSTON ---> 57 HAS_LAPACK64 = numpy.linalg._umath_linalg._ilp64 58 59 _OLD_PROMOTION = lambda: np._get_promotion_state() == 'legacy'

AttributeError: module 'numpy.linalg._umath_linalg' has no attribute '_ilp64'

UPDATE: I've tested the solution given in https://github.com/amaiya/onprem/issues/4 but it solved the problem for the local .py version, and it's too long with no GPU. Therefore I'm still in trouble with the Jupiter notebook version.

Thank you Julien

[amaiya]

Hello,

I just tested the Google Colab notebook to which you are referring and everything worked perfectly. If you downloaded the notebook and running it locally, it may be a problem with your numpy installation. See this post for a fix (which involves uninstalling and reinstalling numpy).

[amaiya]

Also, if you're running the notebook within Google Colab with a GPU, please make sure you restart the notebook, as instructed in the notebook. Feel free to respond to this thread with further questions.

[Velcin]

You're perfectly right. I didn't understand that I must restart the environment, I guess it's because it sounds weird. But all is ok now, sorry for spending your time. Julien