feat!: support v0.11.1

[ILikeIneine]

Purpose

This PR is for supporting vllm v0.11.1

Test Plan

Test Result

(Optional) Documentation Update

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

[gemini-code-assist]

Code Review

This pull request updates the codebase to support vllm v0.11.1, which involves significant refactoring around memory allocation, platform integration, and attention mechanisms. The changes appear to align with the goal of supporting the new vllm version. I have found one critical issue in the device allocator patch that could lead to a runtime error and have provided a fix.

[gemini-code-assist]

There is a potential AttributeError here. The self._sleep_saved_buffers attribute is only initialized within the sleep method, and only when level == 2. If wake_up is called after sleep(level=1) or before any call to sleep, self._sleep_saved_buffers will not exist on the object, causing a crash when len() is called on it.

To prevent this, you should safely check for the attribute's existence before trying to access it.

Suggested change

	if len(self._sleep_saved_buffers):
	model = self.model_runner.model
	for name, buffer in model.named_buffers():
	if name in self._sleep_saved_buffers:
	buffer.data.copy_(self._sleep_saved_buffers[name].data)
	self._sleep_saved_buffers = {}
	if hasattr(self, "_sleep_saved_buffers") and self._sleep_saved_buffers:
	model = self.model_runner.model
	for name, buffer in model.named_buffers():
	if name in self._sleep_saved_buffers:
	buffer.data.copy_(self._sleep_saved_buffers[name].data)
	self._sleep_saved_buffers = {}