[main][feature] Support quarot for eagle3 without embedding (#7038)

### What this PR does / why we need it?
If some `eagle3` model without embed_tokens works with `quarot` target
model, the acceptence rate will drop.
We solve it in this PR.
The relative vllm pr is https://github.com/vllm-project/vllm/pull/36225.

- vLLM main:
4034c3d32e

Signed-off-by: drslark <slarksblood@qq.com>
This commit is contained in:
drslark
2026-03-09 10:43:06 +08:00
committed by GitHub
parent 737dfcf638
commit 6a7115fa0d
5 changed files with 148 additions and 82 deletions

View File

@@ -319,7 +319,7 @@
# https://github.com/vllm-project/vllm/pull/34336
# Future Plan:
# Remove this patch when vLLM merges the PR.
# ** 16. File: worker/patch_qwen3_quarot.py**
# ** 16. File: worker/patch_draft_quarot.py**
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# 1. `vllm.model_executor.models.llama_eagle3.Eagle3LlamaForCausalLM.load_weights`
# Why:
@@ -328,5 +328,7 @@
# How
# Dynamically replace the `load_weights` function at runtime,
# and fix `target_config` into the new implementation with a closure.
# Related PR (if no, explain why):
# https://github.com/vllm-project/vllm/pull/36225
# Future Plan:
# Remove this patch when vLLM merges the PR.