J6E QAT模型训练异常，prepare模型报错

J6E模型QAT训练的时候报错，模型包含backbone，fpn， head多模块，多任务模型，在准备QAT模型的时候prepare报错：

WARNING: check model failed. The common reason is that inputs required for multiple forward are inconsistent. solution:

1. call horizon_plugin_pytorch.utils.check_model.check_qat_model individually

2. rewrite model forward to ensure that inputs of multiple forward are consistent.

（2）输出的信息如下， 是否需要修改？

INFO: cat_19 doesn't support high precisionoutput.

TypeError: rand() received an invalid combination of arguments - got (list, device=torch.device, dtype=QuantDType), but expected one of:

错误如附件，还请帮忙看看， 谢谢！