Xueqing Wu comments

Results 6 comments of


                                            Xueqing Wu

The BERT scores are mismatching while training, so did you use any other script for evaluation.

Hi, what do you mean by the BERT scores are mismatching while training? We do not need to calculate BERT scores during training. As in the README, `bert_score` package is...

How to find the files 'dict.data.txt' and 'dict.text.txt'

Hi, I don't think you need these two files? I think you only need ${BART_DIR}/dict.txt

The BERT scores are mismatching while training, so did you use any other script for evaluation.

@harryniuby 你可以尝试用我发布的数据训练一个模型，然后用你的文本infer。但是我不确定效果怎么样

batch normalization

I also feel confused about this... It seems that as x.size(1) (that is, maximum graph size of a batch) varies from time to time, the size of BatchNorm cannot be...

Fused mlp causes assertion error

Same problem. Disabling fused_mlp works for me. Note: use .pt file, not .safetensors; for some reason .safetensors still triggers the error

NETWORK ERROR DUE TO HIGH TRAFFIC. PLEASE REGENERATE OR REFRESH THIS PAGE. (error_code: 1)

Hi, for me the problem is resolved by running gradio demo **AFTER** the backend is setup, rather than in the second step right after controller is launched as in README....