infer

[zaoruis]

I've finished training on my own data set. However, when I implemented the inference function, it showed that the keys were true. May I ask why? How to solve this problem?

[zaoruis]

I can load the weight of the model after fine tuning and train again, why can't I load the weight of the model after fine tuning for reasoning? In addition, why can the weights you publish be inferred, but the weights I fine-tune cannot be inferred?

[LiheYoung]

The models are stored in state_dict["model"], you should use torch.load(...)["model"] instead.