Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

MPR_Net单卡训练loss下降不明显,且psnr一直是inf,而ssim几乎恒定不变 #565

Open
hhxdestiny opened this issue Jan 19, 2022 · 2 comments
Assignees

Comments

@hhxdestiny
Copy link

hhxdestiny commented Jan 19, 2022

MPR_Net deraining单卡训练loss下降不明显,且psnr一直是inf,而ssim几乎恒定不变。但是原作的MPR可以正常训练,几乎同样的配置,不确定是PaddleGAN/ppgan/metrics/psnr_ssim.py/calculate_psnr.py:96calculate_psnr有问题还是paddle的问题。

运行环境:AI Studio最新版本GPU32G(BML Codelab)
paddle-gpu: develop2.2.2
代码:PaddleGAN最新版本(develop v2.1.0)
配置文件(mprnet_deraining.yaml):

total_iters: 100000
output_dir: output/mprnet
enable_visualdl: True

model:
  name: MPRModel
  generator:
    name: MPRNet
    n_feat: 40
    scale_unetfeats: 20
    scale_orsnetfeats: 16

  char_criterion:
    name: CharbonnierLoss
  edge_criterion:
    name: EdgeLoss

dataset:
  train:
    name: MPRTrain
    rgb_dir: data/mydata/train
    num_workers: 0
    batch_size: 1
    img_options:
      patch_size: 16
  test:
    name: MPRTrain
    rgb_dir: data/mydata/val
    num_workers: 0
    batch_size: 1
    img_options:
      patch_size: 16

lr_scheduler:
  name: CosineAnnealingRestartLR
  learning_rate: !!float 1.25e-5
  periods: [25000, 25000, 25000, 25000]
  restart_weights: [1, 1, 1, 1]
  eta_min: !!float 1e-6

validate:
  interval: 2000
  save_img: false

  metrics:
    psnr: # metric name, can be arbitrary
      name: PSNR
      crop_border: 4
      test_y_channel: True
    ssim:
      name: SSIM
      crop_border: 4
      test_y_channel: True

optimizer:
  name: Adam
  # add parameters of net_name to optim
  # name should in self.nets
  net_names:
    - generator
  beta1: 0.9
  beta2: 0.999
  epsilon: 1e-8

log_config:
  interval: 50
  visiual_interval: 200

snapshot_config:
  interval: 2000

运行过程忘记截图了

@freeman1980
Copy link

inf大概是因为验证集里恰好截图到了近乎纯色的,导致去雨后和原来完全一样,loss不下降我也遇到了,还有loss突然上升到很大,然后就不下来了

@jerrywgz
Copy link
Collaborator

您好,请问问题还是否需要解决,目前相关图像生成能力集成在PaddleMIX中,https://github.com/PaddlePaddle/PaddleMIX/tree/develop 可以在这个repo下提出您的需求

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants