OnnxMatMul4Quantizer: Suppress quantizer logs, no tmp_dir

[jambayk]

Describe your changes

MatMul4BitsQuantizer uses logging.basicConfig to set level to INFO and prints very verbose logs at INFO level. Suppress these logs by manually setting the loggers level to ERROR.
Update ort version requirement to >=1.16.2 since the quantizer will be added to 1.16.2.

Remove the save to tmp_dir -> load -> save steps since it is not needed. We only need to sort the model topologically before saving the model to file. Refer to this discussion for more context #641 (comment).

llama.py: Correct description for --only_config option. Create a new user_script file for the workflow instead of updating the original one.

Checklist before requesting a review

• Add unit tests for this change.
• Make sure all tests can pass.
• Update documents if necessary.
• Format your code by running pre-commit run --all-files
• Is this a user-facing change? If yes, give a description of this change to be included in the release notes.

(Optional) Issue link

[guotuofeng]

Actually, the ort logging severity only impact the logging in c++ code. when I tried last week, I have to use getLogger().setLevel to change the root logger verbosity to make sure the messages could be logged out.

so, do we need change the root logger level so to output the ORT python messages?

[jambayk]

Yes, I looked at the ort logging severity only works for c++.
I was thinking whether we would want (in the future) to set the same logging level for ort python logging by setting the logger for the ort modules at the same level. If we want to do that, we have to store the logging level somewhere. maybe an environment variable.

We don't need to change the root logger. Just need to change the level for the logger for the given name space.
I don't think it is recommended for packages to do anything to the root logger or use basic config. The ort package should not have done that too.

[guotuofeng]

In this week, I have to manually set the root logger level to troubleshoot why the quant_pre_process silently fail(the process exit silently). We need find way to config the ORT python logger level from troubleshooting point of view.

[jambayk]

This sounds good to me. If the logger is created using logging.getLogger(__name__), it makes it easier for us since we could do logging.getLogger("onnxruntime").setLevel(...) to set it for ort.
Or else, I don't know what else we could do without touching the root logger/

[guotuofeng]

Actually, I tried to logging.getLogger("onnxruntime").setLevel(...) but the log is not shown correctly. I don't spend time to investigate the root cause.