I am unable to obtain the results you presented.

[zjunlp]

I am unable to obtain the results you presented.

[njcx-ai]

First of all, your training results deviate too much from the results of the paper. Considering that there is a problem in the training process, it is recommended that you re-download the code to check the training process.

Moreover, there are two more suggestions for this issue:

To fine-tune the RoBERTa-large model, a larger batch size may help you get higher F1-score.
We use 4 GPUs with models NVIDIA GeForce RTX 3090 in training time. Thus the hyper-parameters tuning may be necessary to reproduce the result in your configurations.
Since our model and default setting of hyper-parameters is friendly to the BERT-base and RoBERTa-base model fine-tuning, it's more efficient to reproduce the result on the X-base models. We suggest you try this way.
Thx!

[WingRainSpring]

First of all, your training results deviate too much from the results of the paper. Considering that there is a problem in the training process, it is recommended that you re-download the code to check the training process.

Moreover, there are two more suggestions for this issue:

To fine-tune the RoBERTa-large model, a larger batch size may help you get higher F1-score. We use 4 GPUs with models NVIDIA GeForce RTX 3090 in training time. Thus the hyper-parameters tuning may be necessary to reproduce the result in your configurations. Since our model and default setting of hyper-parameters is friendly to the BERT-base and RoBERTa-base model fine-tuning, it's more efficient to reproduce the result on the X-base models. We suggest you try this way. Thx!

Thank you for your reply. After my testing, I found that there was a problem when dealing with the OOM issue, which caused a significant difference in results. It has now been fixed.