FAQ
This is an ongoing list of issues people sometimes encounter while using Beam. If you’re having an issue, check this list first.
Unable to locate config file
This happens when your current Python environment is different from the Python path Beam was installed in. Often happens while working in virtualenvs.
Solution
- Check which Python environment you’re working in:
which python
. This will print something like/Users/eli/miniconda3/bin/python
. - Install Beam in that environment:
[YOUR PYTHON PATH] -m pip install --upgrade beam-sdk
- Validate the installation by running
beam
Could not load dynamic library
Running Tensorflow models will sometimes result in the following error:
Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /usr/lib/x86_64-linux-gnu:/usr/lib/worker/x86_64-linux-gnu:/usr/local/nvidia/lib64:/usr/local/cuda-12.2/targets/x86_64-linux/lib:$LD_LIBRARY_PATH
Solution
Add LD_LIBRARY_PATH=/usr/local/lib/python3.10/dist-packages/nvidia/cudnn
to your Beam handler function.
Could not load library libcudnn
Sometimes Whisper and other speech models will produce the following error:
Could not load library libcudnn_ops_infer.so.8. Error: libcudnn_ops_infer.so.8: cannot open shared object file: No such file or directory
Please make sure libcudnn_ops_infer.so.8 is in your library path!
Solution
Add this to your commands
field in beam.App.Runtime.Image
:
commands=[
"apt-get update && apt-get install wget"
" && wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.1-1_all.deb",
"dpkg -i cuda-keyring_1.1-1_all.deb",
"lsb_release -a && apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y ffmpeg libpq-dev cuda=11.8.0-1 libcudnn8=8.9.2.*-1+cuda11.8",
]
Using Poetry with Beam
We don’t currently support importing requirements from pyproject.toml
.
Solution
If you’re using Poetry, you should export your requirements to a requirements.txt
and link the path in your python_packages
field in beam.App.Runtime.Image
.
Unable to package app
When you deploy an app to Beam, we create a .zip
folder with the files in your current working directory. This .zip
is limited to 500Mb.
In practice, deployment packages should be much smaller than 500Mb.
Solution
This error typically occurs when you haven’t added a .beamignore
file to your project. Adding a .beamignore
will prevent unneeded files from being included in your deployment package.
Invalid JSON Offset
This error is raised by the JSON parser. This sometimes happens when your JSON syntax is incorrect.
Solution
If you see this error, it’s best to double-check your JSON syntax through a JSON formatter, like this one.
Was this page helpful?