LLaMA-Factory 使用LLaMAFactory SFT Lora 训练MiniCPM-o-2_6的过程中没有问题，但在验证的过程中报错

Reminder

[x] I have read the above rules and searched the existing issues.

System Info

[rank3]: Traceback (most recent call last): [rank3]: File "/data/zhengtianlong/paper02/LLaMA-Factory/src/llamafactory/launcher.py", line 23, in [rank3]: launch() [rank3]: File "/data/zhengtianlong/paper02/LLaMA-Factory/src/llamafactory/launcher.py", line 19, in launch [rank3]: run_exp() [rank3]: File "/data/zhengtianlong/paper02/LLaMA-Factory/src/llamafactory/train/tuner.py", line 103, in run_exp [rank3]: _training_function(config={"args": args, "callbacks": callbacks}) [rank3]: File "/data/zhengtianlong/paper02/LLaMA-Factory/src/llamafactory/train/tuner.py", line 68, in _training_function [rank3]: run_sft(model_args, data_args, training_args, finetuning_args, generating_args, callbacks) [rank3]: File "/data/zhengtianlong/paper02/LLaMA-Factory/src/llamafactory/train/sft/workflow.py", line 127, in run_sft [rank3]: predict_results = trainer.predict(dataset_module["eval_dataset"], metric_key_prefix="predict", **gen_kwargs) [rank3]: File "/data/zhengtianlong/anaconda3/envs/aaai_lmfa/lib/python3.10/site-packages/transformers/trainer_seq2seq.py", line 261, in predict [rank3]: return super().predict(test_dataset, ignore_keys=ignore_keys, metric_key_prefix=metric_key_prefix) [rank3]: File "/data/zhengtianlong/anaconda3/envs/aaai_lmfa/lib/python3.10/site-packages/transformers/trainer.py", line 4151, in predict [rank3]: output = eval_loop( [rank3]: File "/data/zhengtianlong/anaconda3/envs/aaai_lmfa/lib/python3.10/site-packages/transformers/trainer.py", line 4267, in evaluation_loop [rank3]: losses, logits, labels = self.prediction_step(model, inputs, prediction_loss_only, ignore_keys=ignore_keys) [rank3]: File "/data/zhengtianlong/paper02/LLaMA-Factory/src/llamafactory/train/sft/trainer.py", line 114, in prediction_step [rank3]: loss, generated_tokens, _ = super().prediction_step( [rank3]: File "/data/zhengtianlong/anaconda3/envs/aaai_lmfa/lib/python3.10/site-packages/transformers/trainer_seq2seq.py", line 333, in prediction_step [rank3]: generated_tokens = self.model.generate(**generation_inputs, **gen_kwargs) [rank3]: File "/data/zhengtianlong/.cache/huggingface/modules/transformers_modules/MiniCPM-o-2_6/modeling_minicpmo.py", line 786, in generate [rank3]: assert len(input_ids) == len(pixel_values) [rank3]: TypeError: object of type 'NoneType' has no len()

微调数据使用的mllm_demo多模态数据

transformers ==4.48.3 LLama factory ==0.9.3.dev0 torch == 2.4.0+cu124

此外，如果使用transformers 4.47.1 / 4.49.0 / 4.45.0 则不能进行训练