Enabled BnB NF4 inference on Gaudi #20172
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @rsshaik1, 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 introduces support for bitsandbytes NF4 quantization on Gaudi, enhancing the inference capabilities of vLLM. It includes comprehensive tests to validate the functionality and ensures correct device placement for optimal performance. The registration of the custom operator has also been updated to accommodate different platforms.
Highlights
- Gaudi Support: Enabled bitsandbytes NF4 quantization inference on Gaudi.
- Inference Tests: Added inference tests for bitsandbytes NF4 quantization, covering both pre-quantized models and models quantized at runtime.
- Device Placement: Ensured that weights are loaded onto the appropriate device (CUDA or HPU) for bitsandbytes operations.
- Custom Operator Registration: Modified the registration of the
apply_bnb_4bitcustom operator to include a dispatch key for the current platform.
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
This pull request successfully enables BitsAndBytes NF4 quantization for Gaudi (HPU) devices, which is a significant step towards broader hardware support. The changes to the bitsandbytes_loader.py and bitsandbytes.py files correctly adapt the quantization logic and custom operation dispatching for HPU. The new test file test_bitsandbytes_hpu.py introduces specific tests for this functionality. However, the test suite has some areas for improvement, including an incorrect docstring reference, the use of hardcoded local model paths, and a critical issue where the is_quant_method_supported utility does not properly detect HPU support, potentially leading to inaccurate test results. Addressing these points will enhance the robustness and maintainability of the test suite.
|
👋 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 🚀 |
@jeejeelee Yes, we support only NF4 inference on Gaudi. Also, we will delete the HPU specific test file which got added to this PR by mistake (FYI @rsshaik1) |
178b6f5 to
9e4f84a
Compare
Considering that vllm now also supports load_8bit, please add a relevant log to inform users that HPU currently does not support 8-bit inference. |
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
@jeejeelee added the hpu specific assertions |
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai> Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai> Signed-off-by: avigny <47987522+avigny@users.noreply.github.com>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai> Signed-off-by: x22x22 <wadeking@qq.com>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai> Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai> Signed-off-by: Paul Pak <paulpak58@gmail.com>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai> Signed-off-by: Diego-Castan <diego.castan@ibm.com>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
Signed-off-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
This PR adds inference tests for bitsandbytes NF4 quantization on Gaudi, supporting both pre-quantized models and models that are quantized at runtime.