Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

pre-training model #12

Open
lcx0329 opened this issue Aug 3, 2023 · 4 comments
Open

pre-training model #12

lcx0329 opened this issue Aug 3, 2023 · 4 comments

Comments

@lcx0329
Copy link

lcx0329 commented Aug 3, 2023

Excuse me, is the downloaded pre-training model already trained? Why is the test result shown in the figure below?
image

@j-min
Copy link
Owner

j-min commented Aug 3, 2023

I don't have enough information to guide this just from the screenshot, but maybe you can first check the generated outputs? Try colab demo and see if your generation results are similar to the generation results at colab. The colab is based on the pretrained checkpoint.

@lcx0329
Copy link
Author

lcx0329 commented Aug 4, 2023

Due to a mismatch between the PL library and Pytorch, I upgraded the PL library, which resulted in some code changes that needed to be made. After the changes were made, the results are shown in the figure. Here are the modifications I made, and I would like to ask if I have made any mistakes
image
The above is the modification I made. Could you please check if it was due to the above modifications? Thank you

@j-min
Copy link
Owner

j-min commented Aug 4, 2023

I don't see any obvious error from the screenshot, but I don't have information about all the changes you might have made to give useful comments on your case.
I suggest starting from the current pytorch / lightning version in this repo so that you have a clean starting point, and gradually changing one thing at a time to spot what makes the performance drop.

@lcx0329
Copy link
Author

lcx0329 commented Aug 4, 2023

image
The above evaluation results, I don't know why every sentence is incomplete

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants