kongxudaoren

Results 4 comments of kongxudaoren

Dear w5688414 or LeeSureman: i used the code if(args.train): trainer.train() saver = ModelSaver("./cache/model_ckpt_100.pkl") saver.save_pytorch(model,param_only=False) if(args.predict): model=ModelLoader.load_pytorch_model("./cache/model_ckpt_100.pkl") tester = Tester(datasets['test'], model, batch_size=args.test_batch, metrics=metrics) tester.test() But it has error, which said TypeError:...

确实改了,上面的只是先打印出来的,您在后面有个判断,如果是weibo,则dropout都是0.1,确实是这么用的,您还用了什么不同的参数么,hidden怎么是113

dev是56,test更低,只有52点多,调了几次了,一直调不好。楼主您如果有时间提供个log

是bio,楼主使用的bmes么