Yuxiang Zhou

Same Issue, does anyone know how to solve that?

> @ZhouyxNLP Did that resolve? I retrained the whole thing and it didn't happen. As you know threading errors can be such. @cmkumar87 No, I didn't. I retrain many times,...

> @ZhouyxNLP Did you try running it Pytorch 3.6? You may have to update a lit bit of syntax in PreSumm/src/prepro/data_builder.py to port the legacy code to recent convention. But...

> > > @ZhouyxNLP Did you try running it Pytorch 3.6? You may have to update a lit bit of syntax in PreSumm/src/prepro/data_builder.py to port the legacy code to recent...

> I actually succcessfully installed flash-attn 2.5.7 with vllm 0.4.1 and it can be detected with vllm (Using FlashAttention backend). But the performance is remaining the same (there is not...

Hi, thanks for your interest in our work. You can put the GCN encoder between line 851 and line 852 in the [modeling_bert.py](https://github.com/zyxnlp/ACLT/blob/main/examples/aclt/modeling_bert.py) script, and use the structured_output as the...

> 你好，我想用Roberta去替换bert，但是遇到一个错误，不知道什么原因，您能帮我看一下吗？不甚感激！ > Hi thanks for your interest. If you want to use RoBERTa instead of BERT as a pre-trained model, you first need to download the model from https://huggingface.co/roberta-base/tree/main,...