[Hardware][NV] Add support for ModelOpt static scaling checkpoints.

[pavanimajety]

This change adds support for a new quantization class that has the
ability to load ModelOpt checkpoints through HF. It can be invoked via -

llm = LLM(model=model_path, quantization="modelopt")

The checkpoint needs a hf_quant_config.json for loading the right
quantiztion format.

PR Checklist (Click to Expand)

Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.

PR Title and Classification

Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:

• [Bugfix] for bug fixes.
• [CI/Build] for build or continuous integration improvements.
• [Doc] for documentation fixes and improvements.
• [Model] for adding a new model or improving an existing model. Model name should appear in the title.
• [Frontend] For changes on the vLLM frontend (e.g., OpenAI API server, LLM class, etc.)
• [Kernel] for changes affecting CUDA kernels or other compute kernels.
• [Core] for changes in the core vLLM logic (e.g., LLMEngine, AsyncLLMEngine, Scheduler, etc.)
• [Hardware][Vendor] for hardware-specific changes. Vendor name should appear in the prefix (e.g., [Hardware][AMD]).
• [Misc] for PRs that do not fit the above categories. Please use this sparingly.

Note: If the PR spans more than one category, please include all relevant prefixes.

Code Quality

The PR need to meet the following code quality standards:

• We adhere to Google Python style guide and Google C++ style guide.
• Pass all linter checks. Please use format.sh to format your code.
• The code need to be well-documented to ensure future contributors can easily understand the code.
• Include sufficient tests to ensure the project to stay correct and robust. This includes both unit tests and integration tests.
• Please add documentation to docs/source/ if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.

Notes for Large Changes

Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with rfc-required and might not go through the PR.

What to Expect for the Reviews

The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:

• After the PR is submitted, the PR will be assigned to a reviewer. Every reviewer will pick up the PRs based on their expertise and availability.
• After the PR is assigned, the reviewer will provide status update every 2-3 days. If the PR is not reviewed within 7 days, please feel free to ping the reviewer or the vLLM team.
• After the review, the reviewer will put an action-required label on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.
• Please respond to all comments within a reasonable time frame. If a comment isn't clear or you disagree with a suggestion, feel free to ask for clarification or discuss the suggestion.

Thank You

Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!

[robertgshaw2-neuralmagic]

With the utilities that I made for fp8, all of the core functionality of fp8 is now refactored out of fp8.py - let me get that merged and you can build on it.

With the utilities, I made, this PR should not need to touch fp8.py. Instead, you can use the factored out utils withModelOptConfig and ModelOptFp8LinearMethod that mirror the logic we used for AutoFP8. Thus, ModelOptFp8LinearMethod can just handle the weight loading associated with the integration + reuse the utilites we have for connecting to the cutlass kernels

[robertgshaw2-neuralmagic]

Utilities are in and this PR should now be unblocked

[pavanimajety]

@robertgshaw2-neuralmagic @simon-mo @youkaichao Please review 🙏

[robertgshaw2-neuralmagic]

@dsikka going to review

[mgoin]

I believe you should check the config matches here

[pavanimajety]

Config doesn't record "static" or "dynamic" with Model Optimizer config at the moment since it is only relevant for static quantization:

{
    "producer": {
        "name": "modelopt",
        "version": "0.17.0"
    },
    "quantization": {
        "quant_algo": "FP8",
        "kv_cache_quant_algo": null
    }
}

Removing references to "static" since it doesn't look relevant for this quantization class.

[pavanimajety]

/ready

[mgoin]

This looks good enough for now to land! Looking forward to a model checkpoint to test. You may want to merge with main to get the CI to pass

[mgoin]

Just leaving a note that this model doesn't exist yet

[dsikka]

LGTM