[Bug]: assert len(indices) == len(inputs) with Qwen/Qwen2-VL-2B-Instruct

[sssunXw]

Reminder

• I have read the README and searched the existing issues.

System Info

• llamafactory version: 0.9.1.dev0
• Platform: Linux-4.19.24-7.34.cbp.al7.x86_64-x86_64-with-glibc2.35
• Python version: 3.10.14
• PyTorch version: 2.4.0+cu121 (GPU)
• Transformers version: 4.46.1
• Datasets version: 3.0.1
• Accelerate version: 1.0.1
• PEFT version: 0.12.0
• TRL version: 0.9.6
• GPU type: NVIDIA A10
• DeepSpeed version: 0.15.4
• Bitsandbytes version: 0.44.1
• vLLM version: 0.6.3.post1

Reproduction

Using webui or vllm api will result in an error. The startup command is as follows:
CUDA_VISIBLE_DEVICES=0,1,2,3 API_PORT=8000 llamafactory-cli api
--model_name_or_path /mnt/workspace/models/qwen2_vl_7b
--template qwen2_vl
--infer_backend vllm
--vllm_enforce_eager

or llamafactory-cli webui

curl http://localhost:8000/v1/chat/completions
-H "Content-Type: application/json"
-d '{
"model": "Qwen2-VL-7B-Instruct",
"messages": [
{"role": "system", "content": "你是资深分析师."},
{"role": "user", "content": [
{"type": "image_url", "image_url": {"url": "xxx"}},
{"type": "text", "text": "描述用户文化程度水平分布"}
]}
]
}'

Expected behavior

No response

Others

No response