CrossEntropyLoss_LSR is not used?

[huntzhan]

Hi,

Thanks for the great work. Just notice that the CrossEntropyLoss_LSR is not referenced at all. So...does that mean I need to manually replace this line? https://github.com/songyouwei/ABSA-PyTorch/blob/e4da01e1fa1fa57b5bace4cc209a45a3a73521ab/train.py#L152

BTW, I notice there are some reproducible issues as mentioned in #38. I think it would be great if you could provide the actual commands (random seeds, hyperparameters, etc.) for experiments.

[songyouwei]

So...does that mean I need to manually replace this line?

Yes, just replace nn.CrossEntropyLoss with CrossEntropyLoss_LSR.

I think it would be great if you could provide the actual commands (random seeds, hyperparameters, etc.) for experiments.

Among the models implemented, a small part of them provide complete set of hyperparameters, while others do not. So this repo mainly focuses on using a unified code structure to implement all these models, which may sacrifice some consistency in detail. Usually, with default parameters, we can get results close to those in papers.

[songyouwei]

Because the public datasets of ABSA are very small and there is no official segmentation of devset and testset , the experiment results may fluctuate in a certain range. #32 provides a more appropriate way of comparision.

[huntzhan]

thank you.