Error running haystack: Protobuf issue

[sohanasarah]

My working models suddenly started throwing the following error:

RuntimeError: Failed to import transformers.models.rag.modeling_rag because of the following error (look up to see its traceback):
Descriptors cannot not be created directly.
If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.
If you cannot immediately regenerate your protos, some other possible workarounds are:

1. Downgrade the protobuf package to 3.20.x or lower.
2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).

More information: https://developers.google.com/protocol-buffers/docs/news/2022-05-06#python-updates

How to fix that?

[ZanSara]

Hey @sohanasarah we're going to need more context to help you out 🙂 Are you on Colab, on a notebook, or on a local machine? How did you install Haystack and on which OS? Which models are throwing this error?

[sohanasarah]

@ZanSara , I reinstalled Haystack and it resolved the issue.

[vblagoje]

@sohanasarah have you resolved this issue? I suspect that this issue is related to the protobuf package being incompatible with the generated proto files used in transformers. Have you tried a fresh env install of the latest haystack? Please let us know how to recreate this issue. Thanks in advance.

[sohanasarah]

@vblagoje I reinstalled Haystack and it resolved the issue

[sureshvytla]

Team, I'm having the same issue, here are the details if it helps to understand the problem.

Use Case: Semantic Search component consumes messages from Kafka and adds documents to the document store

• Kafka is used for interprocess communication and publishing and consuming messages from a topic using Protobuff
• Consume messages from a topic, converts ProtoBuffer message to Pydantic Model
• Index Pipeline preprocessor converts Pydantic model to Document object (Content and Meta) and adds this to document store
• ElasticSearch used for DocumentStore
• EmbeddingRetriever used for DocumentRetirever

Environment: MacOS
Service deployed with FastAPI and uvicorn

Requirements

farm-haystack[elasticsearch]==1.18.0
fastapi==0.98.0
numpy==1.22.3
pandas==1.4.2
psutil==5.9.1
pynvml==11.5.0
starlette
uvicorn==0.17.6
aiokafka==0.8.1
protobuf==4.23.3

Issue

RuntimeError: Failed to import transformers.pipelines because of the following error (look up to see its traceback):
Descriptors cannot not be created directly.
If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.
If you cannot immediately regenerate your protos, some other possible workarounds are:
 1. Downgrade the protobuf package to 3.20.x or lower.
 2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).

Workaround:

1. Downgrade protobuf to 3.20
2. os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"]="python"