[WIP][Model][Kernel][Bugfix] Commits for new MSFT PhiMoE model #7691
Conversation
Those commits include: * New model named PhiMoE. * New A100 inference kernels for PhiMoE FP8 quantization.
|
👋 Hi! Thank you for contributing to the vLLM project. Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge). To run full CI, you can do one of these:
🚀 |
|
@wenxcs Could you please separate the addition of the new model and the new quantization method into separate PRs? The PhiMoE model seems like it should be simple to land and we have existing quantized FusedMoE methods it could use in the immediate-term |
wenxcs
changed the title
|
Hi @mgoin, thanks for your advice. I'm seperating this PR into two PRs.
|
In this PR, we will introduce:
Model
Kernel
Bugfix
Fix a LongRoPE bug on stale kv cache
[Bug Description]
When the len(input) < N and len(input + generated content) > N (N is the original_position_embeddings of the model. For phi models, N=4096), models using LongRoPE will generate garbage from the N-th token. This bug was captured by this community post.
[Root Cause]
In LongRoPE, before the N-th tokens, model uses short factors, after the N-th token, model switch to use long factors. However, after the N-th token, the previous N tokens' kv cache is stale now.
[Fix]
If initial input < N, whatever number of tokens to generate in this call, keep using short factor.
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,LLMclass, 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:
format.shto format your code.docs/source/if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.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-requiredand 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:
action-requiredlabel on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.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!