Cannot Import Name 'runtime_version' from 'google.protobuf'

[pli-pcln]

Hello, my current meridian version is 1.5.1 and I saw a warning when running model.load_mmm() that it will be deprecated in future releases and instead I should use meridian.schema.serde.meridian_serde.load_meridian.

After installing google-meridian[schema], I get an error when trying to import "from schema.serde import meridian_serde", which is in the title of this thread.

I've looked around online and it mentions upgrading google.protobuf and potentially editing a builder.py. I'm working on a Vertex AI instance so I don't think this is feasible for me. How should I resolve this?

[Shaik-Gouse]

Hello @pli-pcln ,

Thank you for contacting us!

To help me resolve this, could you please provide the full error traceback you receive when you attempt the import? Since my replication in Google Colab was successful using Meridian 1.5.1, the discrepancy likely points to a conflict specific to your Vertex AI instance's pre-installed image. Pinpointing these specific version mismatches will help in identifying the correct environment configuration needed to support the meridian_serde requirements.

Feel free to reach out for any further queries or feedback regarding Meridian.

Google Meridian Support Team

[pli-pcln]

Thanks for your response @Shaik-Gouse

Below is my full error traceback when running from schema.serde import meridian_serde. My version of google.protobuf is 3.20.3
`---------------------------------------------------------------------------
ImportError Traceback (most recent call last)
Cell In[4], line 1
----> 1 from schema.serde import meridian_serde

File ~/.local/lib/python3.10/site-packages/schema/init.py:21
15 """Module containing MMM schema library."""
17 try:
18 # A quick check for schema dependencies.
19 # If this fails, it's likely because meridian was installed without
20 # pip install google-meridian[schema].
---> 21 from mmm.v1.model.meridian import meridian_model_pb2
22 except ModuleNotFoundError as exc:
23 raise ImportError(
24 'Schema dependencies not found. Please install meridian with '
25 'pip install google-meridian[schema].'
26 ) from exc

File ~/.local/lib/python3.10/site-packages/mmm/v1/model/meridian/meridian_model_pb2.py:9
7 from google.protobuf import descriptor as _descriptor
8 from google.protobuf import descriptor_pool as _descriptor_pool
----> 9 from google.protobuf import runtime_version as _runtime_version
10 from google.protobuf import symbol_database as _symbol_database
11 from google.protobuf.internal import builder as _builder

ImportError: cannot import name 'runtime_version' from 'google.protobuf' (/opt/micromamba/lib/python3.10/site-packages/google/protobuf/init.py)`

Let me know if this is helpful.

[Shaik-Gouse]

Hello @pli-pcln ,

Thank you for contacting us!

The ImportError: cannot import name 'runtime_version' from 'google.protobuf' occurs because the mmm-proto-schema package (installed via google-meridian[schema]) generates its Python protobuf files using grpcio-tools >= 1.66.0. These generated files include an import of runtime_version, which was introduced in Protobuf 4.x/5.x. However, your current version, protobuf==3.20.3, is incompatible with these schema files and with google-meridian 1.5.1's primary dependency, tensorflow >= 2.18, which strictly requires protobuf >= 5.28.0.

If you are still facing this issue in your Vertex AI workbench notebook, you can resolve it by running the following command in a notebook cell:

%pip install --upgrade "google-meridian[schema]" protobuf

Crucially, you must restart your Jupyter kernel after the installation finishes. This step is essential to ensure the new version is correctly loaded into memory so the import can succeed.

Feel free to reach out for any further queries or feedback regarding Meridian.

Google Meridian Support Team

[pli-pcln]

@Shaik-Gouse thanks for the reply.

After running this, I notice this will upgrade my version of Meridian to the most updated release, which is 1.5.3. I see in the change log that in this version, support from Python 3.10 will be dropped. My version in my WBI is 3.10.17; will this be a problem?

[Shaik-Gouse]

Hello @pli-pcln ,

Thank you for contacting us!

Your concern regarding Python 3.10 support is valid; support was officially dropped in Meridian 1.5.3 (moving the minimum requirement to Python 3.11) because updated dependencies like JAX 0.7+ are no longer compatible with 3.10, and the version is being deprecated across Vertex AI environments. If possible, the recommended path is to upgrade your Vertex AI Workbench instance to an image running Python 3.11 or 3.12, as this allows you to install the latest Meridian releases and prevents future compatibility issues with other ML libraries. However, if you are currently blocked from upgrading your environment, you can stay on Python 3.10 by pinning your Meridian version while still upgrading protobuf to fix the ImportError by running %pip install "google-meridian[schema]==1.5.1" "protobuf>=5.28.0" in a notebook cell. Regardless of the path you choose, please ensure you restart your Jupyter kernel after the installation finishes so the changes are correctly loaded into memory.

Feel free to reach out for any further queries or feedback regarding Meridian.

Google Meridian Support Team