[Bug]: Running latest docker container fails to find Python extension: libtorch_cuda_cu.so

[radiantone]

Short description of current behavior

Failed to load image Python extension: libtorch_cuda_cu.so: cannot open shared object file: No such file or directory

Video or screenshots

(venv) darren@phoenix:~/PycharmProjects/pyfi$ docker compose up mindsdb
[+] Running 2/0
⠿ Container postgresdb Running 0.0s
⠿ Container mindsdb Recreated 0.1s
Attaching to mindsdb
mindsdb | [nltk_data] Downloading package punkt to /root/nltk_data...
mindsdb | [nltk_data] Unzipping tokenizers/punkt.zip.
mindsdb | /opt/conda/lib/python3.9/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libtorch_cuda_cu.so: cannot open shared object file: No such file or directory
mindsdb | warn(f"Failed to load image Python extension: {e}")
mindsdb |
mindsdb | ✓ telemetry enabled
mindsdb |
mindsdb | Applying database migrations:

Expected behavior

To load Python extension: libtorch_cuda_cu.so

How to reproduce the error

Run the container mindsdb/mindsdb:latest

Anything else?

No response

[legendkartik45]

Can u pls assign this issue to me, will take a look

[radiantone]

@legendkartik45 The admins will have to do it, as I cannot.

[chandrevdw31]

@radiantone I think you will still be able to access the gui via the url http://127.0.0.1:47334

[bilalcorbacioglu]

I'm having the same problem. I have the latest version of `mindsdb/mindsdb'.

a@A-MacBook-Pro ~ % docker run -p 47334:47334 -p 47335:47335 mindsdb/mindsdb


2023-11-21 11:56:01,474 - INFO - generated new fontManager
/opt/conda/lib/python3.9/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libtorch_cuda_cu.so: cannot open shared object file: No such file or directory
  warn(f"Failed to load image Python extension: {e}")
/opt/conda/lib/python3.9/site-packages/flask_restx/namespace.py:116: DeprecationWarning: The body attribute is deprecated, use expect instead
  handle_deprecations(doc)

 ✓ telemetry enabled

[jetjinser]

I also have the same problem 😢.

[elasticcode-ai]

@radiantone I think you will still be able to access the gui via the url http://127.0.0.1:47334

Yeah, but that's not the issue. When running the docker on nvidia hardware you need to be able to use the cuda libraries or training etc will be pathetically slow. So this is an easy fix, the docker just needs to be rebuilt with packages updated. But someone should do it soon since its broke.

[hamishfagg]

@elasticcode-ai @jetjinser @radiantone do you still have this issue with the latest docker images?

[radiantone]

@hamishfagg The original error about the missing shared library (because of outdated torchvision package) is gone, but a few others (warnings only) have shown up now.

mindsdb | [nltk_data] Downloading package punkt to /root/nltk_data...
mindsdb | [nltk_data] Unzipping tokenizers/punkt.zip.
mindsdb | /usr/local/lib/python3.10/site-packages/langchain/init.py:34: UserWarning: Importing Writer from langchain root module is no longer supported. Please use langchain.llms.Writer instead.
mindsdb | warnings.warn(
mindsdb | /usr/local/lib/python3.10/site-packages/torch/cuda/init.py:138: UserWarning: CUDA initialization: The NVIDIA driver on your system is too old (found version 11040). Please update your GPU driver by downloading and installing a new version from the URL: http://www.nvidia.com/Download/index.aspx Alternatively, go to: https://pytorch.org to install a PyTorch version that has been compiled with your version of the CUDA driver. (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:108.)
mindsdb | return torch._C._cuda_getDeviceCount() > 0
mindsdb | 2024-01-03 14:52:53,508 MainProcess INFO mindsdb: ✓ telemetry enabled
mindsdb | /usr/local/lib/python3.10/site-packages/mindsdb/main.py:145: DeprecationWarning: 'JSONEncoder' is deprecated and will be removed in Flask 2.3. Use 'Flask.json' to provide an alternate JSON implementation instead.
mindsdb | db.init()

[ZoranPandovski]

@radiantone This are warnings that you can ignore. Can you successfully access mindsdb on http://127.0.0.1:47334/?

[hamishfagg]

Closing as this doesnt seem to be an issue anymore