What is package “bentoml[io]“? It still has a dependency on ‘pydantic<2’

% pip install --dry-run  "bentoml[io]" | grep pydantic
Collecting pydantic<2 (from bentoml[io])

When installing openllm, it has a dependency on “bentoml[io]” which then downgrades pydantic

pip install --dry-run  openllm==0.4.44 | grep pydantic
Collecting pydantic<2 (from bentoml<1.2,>=1.1.11->bentoml[io]<1.2,>=1.1.11->openllm==0.4.44)

Yet I don’t see many reference to pydantic<2 in any of the bentoml/openllm githubs https://github.com/search?q=org%3Abentoml+%22pydantic%3C2%22&type=code the one exception is https://github.com/bentoml/langchain/blob/76905aa043e3e604b5b34faf5e91d0aedb5ed6dd/docs/api_reference/requirements.txt#L5. The problem for me is that:

pip install  openllm==0.4.44

downgrades pydantic to pydantic-1.10.15 which breaks my environment. Any hints?

@Aaron Pham may have the context for this issue!

a new version of openllm 0.5 is coming out soon to fully support BentoML 1.2 and pydantic. You may try installing from main branch to try it out although it still have some rough edges

0.4.44 is recommended to use with BentoML version < 1.2

Right, we are using bentoml 1.1.11. That doesn’t change the conflict.

% pip install --dry-run bentoml==1.1.11 openllm==0.4.44 |grep pydantic
Collecting pydantic<2 (from bentoml==1.1.11)
  Using cached pydantic-1.10.15-cp311-cp311-macosx_11_0_arm64.whl.metadata (150 kB)
Using cached pydantic-1.10.15-cp311-cp311-macosx_11_0_arm64.whl (2.5 MB)
Would install pydantic-1.10.15

See @Aaron Pham any suggestions?