♻️ fix vllm:main for model_config.task

[prashantgupta24]

It seems model_config.task is deprecated, instead from what I understand we can use model_config.supported_tasks, which is initialized when an llm engine is instantiated:
https://github.com/vllm-project/vllm/pull/21470/files#diff-7eaad0b7dee0626bf29d10081b0f0c5e3ea15a4af97e7b182a4e0d35f8346953R705-R706

To maintain backward compatibility, it's a bit tricky since:

• Earlier version had model_config.task pointing to the task and model_config.supported_tasks as a list of all tasks which could contain more than 1 task

  model_config.task :  generate
  model_config.supported_tasks :  {'embed', 'reward', 'generate', 'classify'}
  

• Latest main now populates model_config.supported_tasks as the only task the model supports.

  model_config.task :  None
  model_config.supported_tasks :  ['generate']
  

[github-actions]

👋 Hi! Thank you for contributing to vLLM support on Spyre.
Just a reminder: Make sure that your code passes all the linting checks, otherwise your PR won't be able to be merged. To do so, first install the linting requirements, then run format.sh and commit the changes. This can be done with uv directly:

uv sync --frozen --group lint --active --inexact

Or this can be done with pip:

uv pip compile --group lint > requirements-lint.txt
pip install -r requirements-lint.txt
bash format.sh

Now you are good to go 🚀

[prashantgupta24]

bot:test
MARKERS="spyre"

[yannicks1]

lgtm

[maxdebayser]

LGTM