YuanEric88 comments

Results 6 comments of


                                            YuanEric88

版本问题

没有添加requirements.txt还挺麻烦的，请问现在知道版本的问题了么？

question about your ChFi-nAnn dataset and bert model

@dolphin-zs Could you show more details on how to use DS-based method to generate labeled data? I am currently working on event extraction for news data, but I am stuck...

论文中的实验结果是使用BERT还是仅用Transformer的呢?

@carrie0307 根据reprod_all_exps.sh可以看出用的是transformer得到的结论。不过我这边复现的时候，在相同的参数设置下，获得的结果要高于论文报告的结果，而且提升幅度挺大的，不知道你是否获得相同结论。

卷积核尺寸

@ZORO-Q 因为在textCNN卷积是一维卷积而非二维，所以在embedding dim上卷积核size与其保持一致。

models/tree/master/research/deep_speech

1.15

Pretrain Dinov2 with custom dataset using Image net 1k weights

Here is the implementation of bicubic interpolation: chkpt = torch.load(cfg.student.pretrained_weights) logger.info(f"OPTIONS -- pretrained weights: loading from {cfg.student.pretrained_weights}") if "pos_embed" in chkpt and chkpt["pos_embed"].shape != student_backbone.pos_embed.shape: logger.info("Positional embedding shape mismatch. Attempting...