[Model] FalconMamba Support #9325

dhiaEddineRhaiem · 2024-10-13T12:11:48Z

This PR adds the support for FalconMamba in VLLM.
In summary , FalconMamba is the same Mamba class without the feed forward block and with the B_C_dt_rms LayerNorm.
Note that FalconMamba is now supported in transformers>=4.45.
I tried with this model https://huggingface.co/tiiuae/falcon-mamba-7b-instruct/ and it worked well locally.
Fixes: #7478

BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE

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.

Adding or changing kernels

Each custom kernel needs a schema and one or more implementations to be registered with PyTorch.

Make sure custom ops are registered following PyTorch guidelines: Custom C++ and CUDA Operators and The Custom Operators Manual
Custom operations that return Tensors require meta-functions. Meta-functions should be implemented and registered in python so that dynamic dims can be handled automatically. See above documents for a description of meta-functions.
Use torch.libary.opcheck() to test the function registration and meta-function for any registered ops. See tests/kernels for examples.
When changing the C++ signature of an existing op, the schema must be updated to reflect the changes.
If a new custom type is needed, see the following document: Custom Class Support in PT2.

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!

github-actions · 2024-10-13T12:11:58Z

👋 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.

🚀

dhiaEddineRhaiem · 2024-10-13T12:13:37Z

vllm/model_executor/layers/layernorm.py

+                 hidden_size: int,
+                 eps: float = 1e-6,
+                 var_hidden_size: Optional[int] = None,
+                 is_learnable: bool = True) -> None:


RMSNorm weights are non learnable for FalconMamba model.
The idea is to add support for non learnable RMSNorm weights, so we can benefit from the same forward types of this class.

Could you explain this a bit more? It seems like this might have been done to work around some issues that popped up during weight loading. Is that right?

And am I right that the weights will always be 1.0 for Falcon Mamba, i.e. we could skip the application of the weights for dt_layernorm, b_layernorm, and c_layernorm?

The idea is to register weights as parameters when they are learnable and register them as buffers whenever they are not so that they will not be included in the state_dict of the model.
the same logic is applied here https://pytorch.org/docs/stable/_modules/torch/nn/modules/normalization.html#RMSNorm (pytorch implementation of RMSNorm)

2.Yes , you are right.

Thanks for the explanation -- I think it would be better to handle this in FalconMambaForCausalLM.load_weights, since it's a special case that only applies to FalconMamba currently.

In load_weights, could you add a condition to check if dt_layernorm, b_layernorm, or c_layernorm is in the name? If this is the case, we can set the weight loader to a function that explicitly sets all of the elements to 1.0, which will make things explicitly clear.

Thanks for the review

I managed to integrate FalconMamba inside mamba.py.

for rmsnorm , i reveretd the changes , but i think there is no need to handle dt_layernorm, b_layernorm, or c_layernorm inside load_weights since they have been initialised as nn.parameters(torch.ones(hidden_size)) inside RMSNorm initial implementation which is compatible with FalconMamba dt,b,c rmsnorms.

tlrmchlsmth

Overall looks good, thanks for the contribution!

I left a few inline comments. Biggest thing for me is to understand what's happening with the layernorm weight to see if this is the right approach.

In a subsequent PR, we should definitely factor out the mixer layer (cc @mzusman if you still plan to tackle it, or I can do it as well)

tlrmchlsmth · 2024-10-13T21:45:32Z

tests/models/decoder_only/language/test_falcon_mamba.py

+
+from ...utils import check_outputs_equal
+
+MODELS = ["tiiuae/falcon-mamba-tiny-dev"]


Instead of adding the test_falcon_mamba.py file, could you add this to the list of models in test_mamba.py?

tlrmchlsmth · 2024-10-13T21:46:15Z

docs/source/models/supported_models.rst

@@ -11,7 +11,7 @@ Text-only Language Models
 ^^^^^^^^^^^^^^^^^^^^^^^^^

 Text Generation
---------------
+--------------- 


nit: remove spurious whitespace

tlrmchlsmth · 2024-10-13T21:46:22Z

docs/source/models/supported_models.rst

@@ -156,7 +161,7 @@ Text Generation
    - Mamba
    - :code:`state-spaces/mamba-130m-hf`, :code:`state-spaces/mamba-790m-hf`, :code:`state-spaces/mamba-2.8b-hf`, etc.
    - ✅︎
-    -
+    -  


ditto, spurious whitespace

tlrmchlsmth · 2024-10-13T21:55:22Z

vllm/model_executor/layers/layernorm.py

+                 hidden_size: int,
+                 eps: float = 1e-6,
+                 var_hidden_size: Optional[int] = None,
+                 is_learnable: bool = True) -> None:


Could you explain this a bit more? It seems like this might have been done to work around some issues that popped up during weight loading. Is that right?

And am I right that the weights will always be 1.0 for Falcon Mamba, i.e. we could skip the application of the weights for dt_layernorm, b_layernorm, and c_layernorm?

tlrmchlsmth · 2024-10-13T21:59:49Z