codeleep
Results
3
comments of
codeleep
> Yes, but it is not that I have not closed the data source of the sharding agent. Instead, the XADataSource and XAConnection used for XA transaction generation cannot be...
> ``` > model = PeftModel.from_pretrained(model, "weights/xxx") > model_dict = model.state_dict() > model_dict.update(torch.load("weights/xxx/adapter_model.bin")) > model.load_state_dict(model_dict) > ``` > > 勉强先这样解决，但是感觉训练时loss算的有问题，我loss一直训到后面全是0了，推理时才能看到效果。 我和 dragononly 老哥出现了一样的问题，大佬有遇到吗