-
https://huggingface.co/docs/transformers/model_doc/bert
I have just got an F1 score of about 88. Is that right? I think the analog result should catch up with the digital ones. |
Beta Was this translation helpful? Give feedback.
Answered by
maljoras
Feb 18, 2024
Replies: 2 comments
-
|
Beta Was this translation helpful? Give feedback.
0 replies
-
|
The analog noise sources will in general reduce the performance of any model (like weight quantization does as well). The reduction can be minimized when the model is (or is made) robust to this noise. See Rasch et al 2023 for a discussion on this (including evaluation on BERT). |
Beta Was this translation helpful? Give feedback.
0 replies
Answer selected by
xsu0960
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The analog noise sources will in general reduce the performance of any model (like weight quantization does as well). The reduction can be minimized when the model is (or is made) robust to this noise. See Rasch et al 2023 for a discussion on this (including evaluation on BERT).