custom_llm_provider is not working with acompletion, but is working with completion

[mrT23]

Addresses the bug described here:
#3480

I am not sure I understand the logic there and why 'completion' and 'acompletion' have a different treatment for the custom_llm_provider parameter, but this solves this specific issue

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
litellm	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	May 6, 2024 4:17pm

[ishaan-jaff]

Hi @mrT23 thanks for the PR - please can you add a test for this scenario

[mrT23]

Hi @mrT23 thanks for the PR - please can you add a test for this scenario

I am not sure I have the technical understanding in litellm to do that.
setting up a real vllm server for testing is hard, and mocking can also be complicated if you don't know every last detail.

[ishaan-jaff]

I am not sure I have the technical understanding in litellm to do that.

I believe editing one of our existing tests would work :

litellm/litellm/tests/test_completion.py

Line 788 in ec63a30

response = await litellm.acompletion(model="claude-2.1", messages=messages)

If you pass custom_llm_provider=anthropic there ^ it would catch this scenario right @mrT23 ?