multi images sft

Thanks for this great work. I want to finetune VideoLLama3 in a scenario where video has been decomposed to multiple images. I construct the dataset jsonl file as followed.
{"image": ["valid_frame/mp4_0.jpg", "valid_frame/mp4_1.jpg", "valid_frame/mp4_2.jpg", "valid_frame/mp4_3.jpg", "valid_frame/mp4_4.jpg"], "conversations": [{"from": "human", "value": "\n"}, {"from": "gpt", "value": "good"}]}
When I finetune this model using stage4 script, I always get this error **IndexError: The shape of the mask [2964] at index 0 does not match the shape of the indexed tensor [14820, 1536] at index 0,** what can I do to solve this issue?