Notice
Recent Posts
Recent Comments
Link
일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | |||||
3 | 4 | 5 | 6 | 7 | 8 | 9 |
10 | 11 | 12 | 13 | 14 | 15 | 16 |
17 | 18 | 19 | 20 | 21 | 22 | 23 |
24 | 25 | 26 | 27 | 28 | 29 | 30 |
31 |
Tags
- straightup
- scowl
- Actor-Critic
- model-free control
- REINFORCE
- non parametic softmax
- sidleup
- thresholding
- Policy Gradient
- noise contrast estimation
- checkitout
- objective functions for machine learning
- MRI
- clip intensity values
- resample
- normalization
- remove outliers
- sample rows
- Excel
- Knowledge Distillation
- domain adaptation
- fastapi
- shadowing
- loss functions
- 3d medical image
- rest-api
- Inorder Traversal
- freebooze
- 자료구조
- pulloff
Archives
- Today
- Total
목록Deep Learning/Seq2Seq (1)
Let's Run Jinyeah
Seq2Seq-스펠링 교정
최근 진행하는 프로젝트에서 단어의 스펠링을 교정해야 하는 일이 생겼다. 이에 대한 해결책으로 기존에 tenosrflow 버전 Deep-Spelling 모델을 pytorch 버전으로 변경해서 사용했다. 모델은 encoder-decoder로 이루어진 seq2seq 모델을 사용해서 pytorch로 seq2seq모델을 구현한 github을 참고했다. 하지만 github에 구현된 seq2seq 모델은 translation이나 question-answer task를 위한 것이기 때문에 모델의 input이 문장이지만 스펠링 교정 task는 모델의 input이 단어이다. 따라서 전처리 부분과 모델 설정 부분을 수정해야 했다. 전처리 모델 인풋 전처리 각 character마다 특정 integer가 할당되는데 학습하는 전..
Deep Learning/Seq2Seq
2020. 11. 27. 05:52