我直接使用inference.py对示例的mp4进行推理发现，尝试了几个mp4文件发现只会回答"!A",请问有什么问题？没报什么其他的错除了”Some weights of HumanOmniQwen2ForCausalLM were not initialized from the model checkpoint at /hdd/u202320081001010/pretrained_model/HumanOmini-7B-Video and are newly initialized: ['model.feature_compressor.vlm_key_projector.bias', 'model.feature_compressor.vlm_key_projector.weight', 'model.feature_compressor.vlm_query_projector.bias', 'model.feature_compressor.vlm_query_projector.weight']“
我直接使用inference.py对示例的mp4进行推理发现，尝试了几个mp4文件发现只会回答"!A",请问有什么问题？没报什么其他的错除了”Some weights of HumanOmniQwen2ForCausalLM were not initialized from the model checkpoint at /hdd/u202320081001010/pretrained_model/HumanOmini-7B-Video and are newly initialized: ['model.feature_compressor.vlm_key_projector.bias', 'model.feature_compressor.vlm_key_projector.weight', 'model.feature_compressor.vlm_query_projector.bias', 'model.feature_compressor.vlm_query_projector.weight']“