[Misc] Suppress duplicated logging regarding multimodal input pipeline

[ywang96]

This message regarding deprecation of legacy multimodal input processing pipeline should really be a print-once info otherwise every request will log it.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[DarkLight1337]

I am hesitant to change the log level to warning since the merged processor isn't fully working yet.

[ywang96]

I am hesitant to change the log level to warning since the merged processor isn't fully working yet.

I can make a new util function print_log_once if that works, otherwise this will make the console output too noisy

[DarkLight1337]

I am hesitant to change the log level to warning since the merged processor isn't fully working yet.

I can make a new util function print_log_once if that works, otherwise this will make the console output too noisy

Sure, that sounds good!

[ywang96]

Confirmed this is working as expected

INFO 11-21 09:49:30 gpu_model_runner.py:515] Loading model weights took 13.1342 GB
INFO 11-21 09:49:32 gpu_executor.py:63] # GPU blocks: 7213
WARNING 11-21 09:49:32 gpu_model_runner.py:549] Skipping CUDA graph capture. Please add -O 3 to use CUDA graphs.
INFO 11-21 09:49:32 preprocess.py:215] Your model uses the legacy input pipeline instead of the new multi-modal processor. Please note that the legacy pipeline will be removed in a future release. For more details, see: https://github.com/vllm-project/vllm/issues/10114
Processed prompts: 100%|████████████████████████████████████████████████████████████████████████| 1000/1000 [00:40<00:00, 24.79it/s, est. speed input: 14773.29 toks/s, output: 1582.72 toks/s]