[Kernel] Add ExLlamaV2 Weight Quantization Support #11348
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
👋 Hi! Thank you for contributing to the vLLM project. 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:
🚀 |
mgoin
reviewed
Dec 20, 2024
There was a problem hiding this comment.
Tried loading a qwen exl2 model and ran into issues with uninitialized parameter usage
vllm serve cgus/Qwen2.5-0.5B-Instruct-exl2 --port 9000 --dtype float16
...
Traceback (most recent call last):
File "/home/mgoin/.local/share/uv/python/cpython-3.12.4-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
self.run()
File "/home/mgoin/.local/share/uv/python/cpython-3.12.4-linux-x86_64-gnu/lib/python3.12/multiprocessing/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/home/mgoin/code/alpin-vllm/vllm/engine/multiprocessing/engine.py", line 368, in run_mp_engine
raise e
File "/home/mgoin/code/alpin-vllm/vllm/engine/multiprocessing/engine.py", line 357, in run_mp_engine
engine = MQLLMEngine.from_engine_args(engine_args=engine_args,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/engine/multiprocessing/engine.py", line 119, in from_engine_args
return cls(ipc_path=ipc_path,
^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/engine/multiprocessing/engine.py", line 71, in __init__
self.engine = LLMEngine(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/engine/llm_engine.py", line 273, in __init__
self.model_executor = executor_class(vllm_config=vllm_config, )
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/executor/executor_base.py", line 36, in __init__
self._init_executor()
File "/home/mgoin/code/alpin-vllm/vllm/executor/gpu_executor.py", line 35, in _init_executor
self.driver_worker.load_model()
File "/home/mgoin/code/alpin-vllm/vllm/worker/worker.py", line 155, in load_model
self.model_runner.load_model()
File "/home/mgoin/code/alpin-vllm/vllm/worker/model_runner.py", line 1094, in load_model
self.model = get_model(vllm_config=self.vllm_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/model_loader/__init__.py", line 12, in get_model
return loader.load_model(vllm_config=vllm_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/model_loader/loader.py", line 364, in load_model
loaded_weights = model.load_weights(
^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/models/qwen2.py", line 506, in load_weights
return loader.load_weights(weights)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/models/utils.py", line 237, in load_weights
autoloaded_weights = set(self._load_module("", self.module, weights))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/models/utils.py", line 198, in _load_module
yield from self._load_module(prefix,
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/models/utils.py", line 175, in _load_module
loaded_params = module_load_weights(weights)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/models/qwen2.py", line 396, in load_weights
weight_loader(param, loaded_weight)
File "/home/mgoin/code/alpin-vllm/vllm/model_executor/model_loader/weight_utils.py", line 531, in default_weight_loader
if param.numel() == 1 and loaded_weight.numel() == 1:
^^^^^^^^^^^^^
File "/home/mgoin/venvs/alpin/lib/python3.12/site-packages/torch/nn/parameter.py", line 168, in __torch_function__
raise ValueError(
ValueError: Attempted to use an uninitialized parameter in <method 'numel' of 'torch._C.TensorBase' objects>. This error happens when you are using a `LazyModule` or explicitly manipulating `torch.nn.parameter.UninitializedParameter` objects. When using LazyModules Call `forward` with a dummy batch to initialize the parameters before calling torch functions
Co-authored-by: Michael Goin <[email protected]>
Co-authored-by: Michael Goin <[email protected]>
|
Thank you for this WIP PR! A quick question: I just built vllm HEAD with this PR applied, and I can't get it to work. Just would like to know if I built it right and it is expected, or I messed up 😅 It first downloads the weights fine, but then I'm getting this error: Thanks 😃 |
|
This pull request has merge conflicts that must be resolved before it can be |
38 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Work in Progress.
FIX #3203