Remove extra tensor on CPU #20693
Remove extra tensor on CPU #20693
Conversation
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
There was a problem hiding this comment.
Summary of Changes
Hello @maxdebayser, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request optimizes the MinPLogitsProcessor by intelligently managing tensor allocation based on the execution device. For CPU-bound operations, it removes the creation and copying of a redundant device tensor, leading to improved memory efficiency and streamlined data handling without impacting GPU performance.
Highlights
- Memory Optimization: I've refactored the
MinPLogitsProcessorto eliminate an unnecessary device tensor when running on CPU ortorch.compile-supported devices. This reduces memory overhead by reusing the existing CPU tensor instead of allocating a duplicate on the device. - Conditional Tensor Management: I've introduced a
double_tensorflag to conditionally manage tensor allocation and data copying. If the target device is CPU, themin_p_devicetensor now directly references themin_p_cpu_tensor, and thecopy_operation is skipped, as it would be redundant.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist is currently in preview and may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments to provide feedback.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
The pull request refactors the MinPLogitsProcessor to avoid using an unnecessary double tensor scheme on the CPU, which was previously used for coalescing data movements on the GPU. The changes introduce a double_tensor flag to conditionally allocate the device tensor based on whether the device is different from the CPU.
vllm/v1/sample/logits_processor.py
Outdated
| dtype=torch.float32, | ||
| device=device) | ||
|
|
||
| self.double_tensor = torch.device("cpu") != torch.device(device) |
There was a problem hiding this comment.
It might be beneficial to add a comment explaining the purpose of the double_tensor variable and why it's needed to check for CPU vs device equality.
|
👋 Hi! Thank you for contributing to the vLLM project. 💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run 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 either: Add 🚀 |
|
Thanks @maxdebayser! This will apply similarly to many of the other sampling parameters. |
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
|
@njhill is it ok to merge? |
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com> Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com> Signed-off-by: avigny <47987522+avigny@users.noreply.github.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com> Signed-off-by: x22x22 <wadeking@qq.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com> Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com> Signed-off-by: Paul Pak <paulpak58@gmail.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com> Signed-off-by: Diego-Castan <diego.castan@ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
On GPU the MinPLogitsProcessor uses a double tensor scheme similar to what the gpu model runner uses to coalesce data movements from the CPU to the GPU. On the CPU or on devices that are supported only via torch.compile such as Spyre, the extra tensor is not necessary.
cc: @afeldman-nm