cooper12121 / DIE-EC

4 stars 0 forks source link

something wrong about WEC-Zh #4

Open eeehco opened 1 month ago

eeehco commented 1 month ago

Is there anything in the code that needs to be modified when training with the WEC Zh dataset? I have encountered some issues that WEC Eng does not have, such as when running 'prepare_embed. py':

image

cooper12121 commented 1 month ago

As for this error, because i use a different pretrained model for wec_zh, so the code has little difference with current code, but I believe it's easy to fix, as well as the error you mentioned. just check the return value of self.model, and get the last_hidden_state in another way. cause I'm not in school recently, i‘ll update this for zh some days later.

eeehco commented 1 month ago

As for this error, because i use a different pretrained model for wec_zh, so the code has little difference with current code, but I believe it's easy to fix, as well as the error you mentioned. just check the return value of self.model, and get the last_hidden_state in another way. cause I'm not in school recently, i‘ll update this for zh some days later.

I have fixed this error. There is now a feature dimension issue in Roberta large's model, which may be due to the pre trained model. I look forward to your updates on the code. Thank you for taking the time out of your busy schedule to reply to me!

image

eeehco commented 1 month ago

As for this error, because i use a different pretrained model for wec_zh, so the code has little difference with current code, but I believe it's easy to fix, as well as the error you mentioned. just check the return value of self.model, and get the last_hidden_state in another way. cause I'm not in school recently, i‘ll update this for zh some days later.

I have fixed this error. There is now a feature dimension issue in Roberta large's model, which may be due to the pre trained model. I look forward to your updates on the code. Thank you for taking the time out of your busy schedule to reply to me!

image

cooper12121 commented 3 weeks ago

As for this error, because i use a different pretrained model for wec_zh, so the code has little difference with current code, but I believe it's easy to fix, as well as the error you mentioned. just check the return value of self.model, and get the last_hidden_state in another way. cause I'm not in school recently, i‘ll update this for zh some days later.

I have fixed this error. There is now a feature dimension issue in Roberta large's model, which may be due to the pre trained model. I look forward to your updates on the code. Thank you for taking the time out of your busy schedule to reply to me!

image

I have upload zh code in brach "zh". now you can use this new branch for chinese process. If there are any problems, feel free to reach out to me.