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 |
- agnoster폰트
- 캐나다 은행계좌 개설
- 콘도렌트
- Generative model
- agnoster폰트꺠짐
- 캐나다월세
- agnoster폰트깨짐
- 캐나다 TD 예약
- 프라이탁
- GenerativeModel
- 캐나다 TD 한국인 예약
- MachineLearning
- Docker
- DTW
- EATS
- GAN
- 딥러닝
- iterm2자동완성
- Flow
- iterm2꾸미기
- 캐나다콘도렌트
- 터미널꾸미기
- 프라이탁 존버
- pytorch
- 머신러닝
- iterm2환경설정
- 캐나다 TD 한국인 직원 예약
- 캐나다 은행 계좌 개설
- Normalizing flow
- 캐나다 TDBAnk
- Today
- Total
TechNOTE
NVIDIA A100 사용하기 본문
회사에 A100이 들어왔대서 써보려구 걍 돌려봤는데..
띠용~~ 7배가 느리다 ;; 당황쓰
찾아보니까.. A100에서 지원하는 TensorFloat-32 (TF32) 를 사용해서 학습을 시켜야 한다고 한다.
fp32 연산을 가지고 10배 빠르게 학습이 가능하고 fp16 을 사용하는 학습보다 훨씬 안정적으로 가능하다니.. 너무좋잖아?
근데 또 찾아 보면 TF32 is supported in the NVIDIA Ampere GPU architecture and is enabled by default.
라고 해서 자동으로 enable 되는 것 같은데.. 도대체 어디서 병목이 생기는거지..
찾아보니까 cudnn 문제라고한다.
https://discuss.pytorch.org/t/a100-is-slower-than-1080ti-with-pytorch/124913
A100 is slower than 1080ti with pytorch
First, I found out that the A100 supports: cuda 11.0 nvidia driver 450 So, when using the A100, I adjusted the environment as follows. cuda 11.0 cudnn 8005 (I checked by torch.backends.cudnn.version()) nvidia driver 450 (more detailed, 450.119.04) pytorch
discuss.pytorch.org
Installation Guide :: NVIDIA Deep Learning cuDNN Documentation
The NVIDIA® CUDA® Deep Neural Network library™ (cuDNN) is a GPU-accelerated library of primitives for deep neural networks. cuDNN provides highly tuned implementations for standard routines such as forward and backward convolution, pooling, normalizati
docs.nvidia.com
여기서 나와있는대로 cudnn 설치하니까 해결되었다
https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html#install-linux
'딥러닝, 연구' 카테고리의 다른 글
[논문리뷰] You Only Look Once:Unified, Real-Time Object Detection (2) | 2021.05.27 |
---|---|
[논문리뷰] Flow ++ (2) | 2021.05.05 |
[GLOW] Affine Coupling Layer (0) | 2021.04.30 |
Set python3 as default in ubuntu (0) | 2021.04.28 |
[논문리뷰] Glow: Generative Flow with Invertible 1×1 Convolutions (2) | 2020.12.02 |