HW8: Constrained Decoding #549
Replies: 4 comments 9 replies
-
|
Do I require to check the coverage in this homework? Or, is it okay to skip like HW 5, 6, and 7? |
Beta Was this translation helpful? Give feedback.
-
|
안녕하세요, 현재 모델이 모든 로짓 후보가 아닌 일부 후보군만 반환하는 것 같습니다. 따라서 Model.generate의 인풋에만 임의로 문자를 추가하여(예: acc + '\n') 특정 토큰(예: ```)을 반환하도록 유도하는 것이 허용되나요? Is it acceptable to intentionally add arbitrary characters (for example, '\n') to the input of Model.generate in order to encourage the model to produce a specific token (such as ```)? |
Beta Was this translation helpful? Give feedback.
-
|
(다른 학우분께서 질문하셨던 부분인데 글이 사라진 것 같아 다시 질문 드립니다.) 안녕하세요, Hello, |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.


Uh oh!
There was an error while loading. Please reload this page.
-
안녕하세요.
마지막 과제 HW8에서 여러분은 조건부 생성 (constrained decoding) 알고리즘을 구현하게 됩니다.
조건부 생성은 언어 모델이 생성한 출력이 주어진 조건을 만족하도록 강제하면서 출력을 생성하는 기술입니다.
아래 링크로 저장소를 생성한 후 과제를 시작해주시기 바랍니다.
https://classroom.github.com/a/77bar4D1
다음은 과제 8의 세부 평가 기준입니다.
과제 마감 기한은 12월 21일(일) 이며, 늦은 제출이 없다는 점에 유의하시기 바랍니다.
감사합니다.
Hello everyone,
For the final assignment, HW8, you will implement a constrained decoding algorithm.
Constrained decoding is a technique that generates output while enforcing that the output produced by a language model satisfies a given constraint.
Please begin the assignment by creating a repository via the link below:
https://classroom.github.com/a/77bar4D1
Below are the detailed grading criteria for HW8:
The deadline for this assignment is Sunday, December 21st. Please note that late submissions will not be accepted.
Beta Was this translation helpful? Give feedback.
All reactions