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
- Ubuntu 설치
- 전체내용
- Lagrange Multiplier
- 듀얼부트
- Git branch 보이기
- DenseNet
- 학습성공
- VGGNet
- stepsize
- #ubuntu 14.04(LTS)
- 3D convolution
- weight histogram
- tensorboard
- log파일
- twolay
- Xpad설치
- canon mf416dw
- #cuda
- Ubuntu동글
- Rust설치
- 포트번호 변경
- restnet
- 리눅스 비밀번호
- #python gpu
- #enviroment variable
- Hidden layer output
- 화면확대고정
- ubuntu 원격
- 치환 기능
- #cudnn
Archives
- Today
- Total
목록weight histogram (1)
save the world
4주차
Model Output Check 학습모델의 최종결과인 l5를 확인하기위해 l5를 numpy.array로 받았는데 몇번의 iteration 후 모든 픽셀의 값이 0으로 나왔다. 원인을 찾기위해 weight 값 (커널)을 확인하려고 tensorboard를 사용했다(아래문단).Tensorboard: Weight Histogram 텐서보드에서 첫번째 레이어의 weight값이 한번 학습했을 때 아주 크게 변하는 것을 확인하고 이를 장기적인 측면에서 확인하고자 텐서보드를 사용하여 그래프로 보고싶었다. weight값을 텐서보드로 넘기기 위해서 모든 weight값을 저장하기위해 tf.summary.histogram("w1_summ", w1) 명령어를 사용하고 실행하였더니 InvalidArgumentError: Nan..
Deep Learning Diary
2018. 2. 6. 17:20