[python] Use vllm chat object

[xyang16]

Description

Brief description of what this PR is about

• If this change is a backward incompatible change, why must this change be made?
• Interesting edge cases to note here

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

Checklist:

• Please add the link of Integration Tests Executor run with related tests.
• Have you manually built the docker image and verify the change?
• Have you run related tests? Check how to set up the test environment here; One example would be pytest tests.py -k "TestCorrectnessLmiDist" -m "lmi_dist"
• Have you added tests that prove your fix is effective or that this feature works?
• Has code been commented, particularly in hard-to-understand areas?
• Have you made corresponding changes to the documentation?

Feature/Issue validation/testing

Please describe the Unit or Integration tests that you ran to verify your changes and relevant result summary. Provide instructions so it can be reproduced.
Please also list any relevant details for your test configuration.

• Test A
  Logs for Test A

• Test B
  Logs for Test B

[siddvenk]

This should be ok in the LMI and Neuron containers (we should validate on the neuron side that this does work), but I don't know if it will work in the trtllm container since we don't install vllm there.

We should either install vllm in the trtllm container, or retain the old messages format for trtllm.

[xyang16]

After some testing, I think we need to retain the old format for trtllm container.

[siddvenk]

is it possible to base this choice of the config option.rolling_batch=x?

[xyang16]

option.rolling_batch may be auto, this will be lmi-dist or trtllm depends on which container it is. So it's hard to tell which rolling batch it is.

[siddvenk]

Maybe we could set a config within the RB class like use_vllm_chat_completions? I think I would prefer that since i'm not sure whether using VllmRollingBatch with Neuron (a valid use case) supports some of the utilities we are using from vllm since we're pulling those neuron's vllm repo

[xyang16]

Sounds good. Added use_vllm_chat_completions()

[siddvenk]

@xyang16 i added 2 small changes to this PR

1. fix the max_new_tokens issue in the integ test client
2. pass all the parsed mm data rather than just images to the request