Reminder

• I have read the above rules and searched the existing issues.

System Info

• llamafactory version: 0.9.4.dev0
• Platform: Linux-5.15.0-119-generic-x86_64-with-glibc2.35
• Python version: 3.11.14
• PyTorch version: 2.9.1+cu128 (GPU)
• Transformers version: 4.57.1
• Datasets version: 4.0.0
• Accelerate version: 1.11.0
• PEFT version: 0.17.1
• GPU type: NVIDIA H200
• GPU number: 8
• GPU memory: 139.81GB
• TRL version: 0.9.6
• DeepSpeed version: 0.16.9
• Git commit: aeda079
• Default data directory: detected

Reproduction

在用sft方法训练qwen3vl时，数据用的是video数据，且是多轮对话，报错idx out of range, 这里代码中是默认video_metadata长度和messgaes长度相同吗，如果多轮对话中user又输入了一个video就不支持吗

Others

No response