From a94189295b8b9c1d952be438b49ed5793db59159 Mon Sep 17 00:00:00 2001 From: weizhoublue <45163302+weizhoublue@users.noreply.github.com> Date: Sun, 17 May 2026 08:54:27 +0800 Subject: [PATCH] Fix Weight loading for Qwen3.5-MTP and Qwen3-VL using runai_streamer (#42716) Signed-off-by: weizhoublue --- vllm/model_executor/models/qwen3_5_mtp.py | 4 ++-- vllm/model_executor/models/qwen3_vl_moe.py | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/vllm/model_executor/models/qwen3_5_mtp.py b/vllm/model_executor/models/qwen3_5_mtp.py index e86b205b9f..5622065a7f 100644 --- a/vllm/model_executor/models/qwen3_5_mtp.py +++ b/vllm/model_executor/models/qwen3_5_mtp.py @@ -175,8 +175,8 @@ class Qwen3_5MultiTokenPredictor(nn.Module): param, curr_expert_weight, name, - shard_id, - expert_id, + shard_id=shard_id, + expert_id=expert_id, return_success=True, ) if success: diff --git a/vllm/model_executor/models/qwen3_vl_moe.py b/vllm/model_executor/models/qwen3_vl_moe.py index 4b38ff9b9e..195b3355e3 100644 --- a/vllm/model_executor/models/qwen3_vl_moe.py +++ b/vllm/model_executor/models/qwen3_vl_moe.py @@ -152,8 +152,8 @@ class Qwen3MoeLLMModel(Qwen3MoeModel): param, curr_expert_weight, name, - shard_id, - expert_id, + shard_id=shard_id, + expert_id=expert_id, return_success=True, ) if success: