Add stargan-v2 style FID calculation. For stargan-v2 FID calculation, we will use an Inception-V3 model weight converted from official torchvision implementation (I have uploaded it to AIStudio: https://aistudio.baidu.com/aistudio/datasetdetail/51890). Note that the original Inception model weights is converted from an unofficial tensorflow model, and generate different result compared to the torchvision one.

add --style command line option to let user choose stargan or gan-compression style (by default gan-compression style will be used).

move with fluid.dygraph.guard() call into fid module for two reason: 1. this inference model didn't work in dygraph mode because it's not dynamic graph model and thus it's faster, so we dynamically choose whether to use dygraph mode after style is determined. (for gan-compression style, the original dygraph version model is used. And for stargan style, this new inference model will be used) 2. easier to use for end user (no need to call fluid.dygraph.guard() explicitly)

整合visualdl

1. 修改了trianer.py，添加了global step、vdl logger对象进行写入。修改visual函数，添加了is_save_image参数控制是否直接写入文件夹，参考min_max添加了image_num控制每次写入的图像数量。
2. 修改了visual.py的tensor2img函数，支持3\4维的图像转换。
3. main.py中添加了异常捕获，ctrl+c时可以自动保存当前数据。

尝试了在AIStudio中运行，遇到了人机验证错误，堆栈溢出的错误，所以想尝试本地运行，但是发现必要的环境是无法安装，如下所示：

➜  PaddleGAN git:(develop) pip install paddlepaddle-gpu==2.1.0.post101 -f https://mirror.baidu.com/pypi/simple
Looking in links: https://mirror.baidu.com/pypi/simple
ERROR: Could not find a version that satisfies the requirement paddlepaddle-gpu==2.1.0.post101
ERROR: No matching distribution found for paddlepaddle-gpu==2.1.0.post101
WARNING: You are using pip version 20.3.1; however, version 21.2.4 is available.
You should consider upgrading via the '/usr/local/opt/[email protected]/bin/python3.9 -m pip install --upgrade pip' command.

所以是我的操作问题吗？我的设备环境如下：

AttributeError: '_dlib_pybind11.rectangle' object has no attribute 'rect'

得到的结果是这样的

rectangles[[(26, 73) (167, 214)], [(229, 73) (370, 214)]]

应该是版本的问题

运行环境：

Ai Studio高级版

前置操作：

1. ! python3 -m pip install paddlepaddle-gpu==2.0.0b0 -i https://mirror.baidu.com/pypi/simple
2. ! git clone https://hub.fastgit.org/PaddlePaddle/PaddleGAN.git
3. cd PaddleGAN && python -u applications/tools/first-order-demo.py --driving_video datasets/unravel.flv --source_image datasets/ssc.jpg --relative --adapt_scale

报错信息

Traceback (most recent call last):
  File "applications/tools/first-order-demo.py", line 18, in <module>
    from ppgan.apps.first_order_predictor import FirstOrderPredictor
  File "/home/aistudio/PaddleGAN/ppgan/apps/__init__.py", line 2, in <module>
    from .deepremaster_predictor import DeepRemasterPredictor
  File "/home/aistudio/PaddleGAN/ppgan/apps/deepremaster_predictor.py", line 24, in <module>
    from ppgan.models.generators.remaster import NetworkR, NetworkC
  File "/home/aistudio/PaddleGAN/ppgan/models/__init__.py", line 20, in <module>
    from .makeup_model import MakeupModel
  File "/home/aistudio/PaddleGAN/ppgan/models/makeup_model.py", line 30, in <module>
    from ..datasets.makeup_dataset import MakeupDataset
  File "/home/aistudio/PaddleGAN/ppgan/datasets/__init__.py", line 15, in <module>
    from .unpaired_dataset import UnpairedDataset
  File "/home/aistudio/PaddleGAN/ppgan/datasets/unpaired_dataset.py", line 4, in <module>
    from .base_dataset import BaseDataset, get_transform
  File "/home/aistudio/PaddleGAN/ppgan/datasets/base_dataset.py", line 10, in <module>
    from .transforms import transforms as T
  File "/home/aistudio/PaddleGAN/ppgan/datasets/transforms/__init__.py", line 1, in <module>
    from .transforms import PairedRandomCrop, PairedRandomHorizontalFlip
  File "/home/aistudio/PaddleGAN/ppgan/datasets/transforms/transforms.py", line 23, in <module>
    TRANSFORMS.register(T.Transpose)
AttributeError: module 'paddle.vision.transforms' has no attribute 'Transpose'

如题，PaddleGAN MPR_Net训练过程中GPU开飞机，显存持续上涨，直到100%。

运行环境：AI Studio最新版本GPU32G（BML Codelab 2.2.1 Python3） 代码：PaddleGAN最新版本（v2.1.0） 训练命令：

python -u tools/main.py --config-file configs/mprnet_test.yaml

配置文件（mprnet_test.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 2e-4
  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

部分训练日志

[01/18 14:28:56] ppgan.engine.trainer INFO: Iter: 950/100000 lr: 1.993e-04 loss: 7.585 batch_cost: 0.13380 sec reader_cost: 0.00008 sec ips: 7.47377 images/s eta: 3:40:53
[01/18 14:29:03] ppgan.engine.trainer INFO: Iter: 1000/100000 lr: 1.992e-04 loss: 10.412 batch_cost: 0.13585 sec reader_cost: 0.00009 sec ips: 7.36111 images/s eta: 3:44:09
[01/18 14:29:09] ppgan.engine.trainer INFO: Iter: 1050/100000 lr: 1.991e-04 loss: 26.876 batch_cost: 0.12891 sec reader_cost: 0.00008 sec ips: 7.75758 images/s eta: 3:32:35
[01/18 14:29:16] ppgan.engine.trainer INFO: Iter: 1100/100000 lr: 1.991e-04 loss: 13.800 batch_cost: 0.14733 sec reader_cost: 0.00010 sec ips: 6.78764 images/s eta: 4:02:50
[01/18 14:29:23] ppgan.engine.trainer INFO: Iter: 1150/100000 lr: 1.990e-04 loss: 13.804 batch_cost: 0.12896 sec reader_cost: 0.00008 sec ips: 7.75449 images/s eta: 3:32:27
[01/18 14:29:29] ppgan.engine.trainer INFO: Iter: 1200/100000 lr: 1.989e-04 loss: 4.678 batch_cost: 0.13077 sec reader_cost: 0.00008 sec ips: 7.64728 images/s eta: 3:35:19
[01/18 14:29:36] ppgan.engine.trainer INFO: Iter: 1250/100000 lr: 1.988e-04 loss: 11.096 batch_cost: 0.13756 sec reader_cost: 0.00008 sec ips: 7.26970 images/s eta: 3:46:23
[01/18 14:29:44] ppgan.engine.trainer INFO: Iter: 1300/100000 lr: 1.987e-04 loss: 24.029 batch_cost: 0.14615 sec reader_cost: 0.00010 sec ips: 6.84228 images/s eta: 4:00:25
[01/18 14:29:51] ppgan.engine.trainer INFO: Iter: 1350/100000 lr: 1.986e-04 loss: 14.311 batch_cost: 0.14573 sec reader_cost: 0.00009 sec ips: 6.86211 images/s eta: 3:59:36
[01/18 14:29:58] ppgan.engine.trainer INFO: Iter: 1400/100000 lr: 1.985e-04 loss: 17.655 batch_cost: 0.14838 sec reader_cost: 0.00010 sec ips: 6.73934 images/s eta: 4:03:50
[01/18 14:30:06] ppgan.engine.trainer INFO: Iter: 1450/100000 lr: 1.984e-04 loss: 9.815 batch_cost: 0.14712 sec reader_cost: 0.00010 sec ips: 6.79716 images/s eta: 4:01:38
[01/18 14:30:13] ppgan.engine.trainer INFO: Iter: 1500/100000 lr: 1.982e-04 loss: 20.609 batch_cost: 0.15017 sec reader_cost: 0.00010 sec ips: 6.65920 images/s eta: 4:06:31
[01/18 14:30:21] ppgan.engine.trainer INFO: Iter: 1550/100000 lr: 1.981e-04 loss: 22.655 batch_cost: 0.14877 sec reader_cost: 0.00010 sec ips: 6.72189 images/s eta: 4:04:06
[01/18 14:30:28] ppgan.engine.trainer INFO: Iter: 1600/100000 lr: 1.980e-04 loss: 22.751 batch_cost: 0.14651 sec reader_cost: 0.00010 sec ips: 6.82556 images/s eta: 4:00:16
[01/18 14:30:35] ppgan.engine.trainer INFO: Iter: 1650/100000 lr: 1.979e-04 loss: 368.427 batch_cost: 0.14723 sec reader_cost: 0.00010 sec ips: 6.79201 images/s eta: 4:01:20
[01/18 14:30:43] ppgan.engine.trainer INFO: Iter: 1700/100000 lr: 1.977e-04 loss: 19.512 batch_cost: 0.14658 sec reader_cost: 0.00009 sec ips: 6.82226 images/s eta: 4:00:08
[01/18 14:30:50] ppgan.engine.trainer INFO: Iter: 1750/100000 lr: 1.976e-04 loss: 9.517 batch_cost: 0.14538 sec reader_cost: 0.00009 sec ips: 6.87872 images/s eta: 3:58:03
[01/18 14:30:57] ppgan.engine.trainer INFO: Iter: 1800/100000 lr: 1.975e-04 loss: 17.127 batch_cost: 0.13771 sec reader_cost: 0.00009 sec ips: 7.26190 images/s eta: 3:45:22
[01/18 14:31:03] ppgan.engine.trainer INFO: Iter: 1850/100000 lr: 1.973e-04 loss: 9.396 batch_cost: 0.13269 sec reader_cost: 0.00009 sec ips: 7.53654 images/s eta: 3:37:03
[01/18 14:31:10] ppgan.engine.trainer INFO: Iter: 1900/100000 lr: 1.972e-04 loss: 11.732 batch_cost: 0.12869 sec reader_cost: 0.00008 sec ips: 7.77042 images/s eta: 3:30:24
[01/18 14:31:17] ppgan.engine.trainer INFO: Iter: 1950/100000 lr: 1.970e-04 loss: 6.371 batch_cost: 0.14135 sec reader_cost: 0.00009 sec ips: 7.07441 images/s eta: 3:50:59
[01/18 14:31:24] ppgan.engine.trainer INFO: Iter: 2000/100000 lr: 1.969e-04 loss: 19.288 batch_cost: 0.14611 sec reader_cost: 0.00009 sec ips: 6.84437 images/s eta: 3:58:38
[01/18 14:31:24] ppgan.engine.trainer INFO: Test iter: [0/836]
/home/aistudio/work/PaddleGAN/ppgan/metrics/psnr_ssim.py:176: RuntimeWarning: Mean of empty slice.
  return ssim_map.mean()
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/numpy/core/_methods.py:188: RuntimeWarning: invalid value encountered in double_scalars
  ret = ret.dtype.type(ret / rcount)
[01/18 14:31:30] ppgan.engine.trainer INFO: Test iter: [50/836]
[01/18 14:31:36] ppgan.engine.trainer INFO: Test iter: [100/836]
[01/18 14:31:41] ppgan.engine.trainer INFO: Test iter: [150/836]
[01/18 14:31:47] ppgan.engine.trainer INFO: Test iter: [200/836]
[01/18 14:31:52] ppgan.engine.trainer INFO: Test iter: [250/836]
[01/18 14:31:58] ppgan.engine.trainer INFO: Test iter: [300/836]
[01/18 14:32:04] ppgan.engine.trainer INFO: Test iter: [350/836]
[01/18 14:32:09] ppgan.engine.trainer INFO: Test iter: [400/836]
[01/18 14:32:15] ppgan.engine.trainer INFO: Test iter: [450/836]
[01/18 14:32:20] ppgan.engine.trainer INFO: Test iter: [500/836]
[01/18 14:32:26] ppgan.engine.trainer INFO: Test iter: [550/836]
[01/18 14:32:31] ppgan.engine.trainer INFO: Test iter: [600/836]
[01/18 14:32:37] ppgan.engine.trainer INFO: Test iter: [650/836]
[01/18 14:32:42] ppgan.engine.trainer INFO: Test iter: [700/836]
[01/18 14:32:48] ppgan.engine.trainer INFO: Test iter: [750/836]
[01/18 14:32:54] ppgan.engine.trainer INFO: Test iter: [800/836]
[01/18 14:32:58] ppgan.engine.trainer INFO: Metric psnr: inf
[01/18 14:32:58] ppgan.engine.trainer INFO: Metric ssim: nan
[01/18 14:33:05] ppgan.engine.trainer INFO: Iter: 2050/100000 lr: 1.967e-04 loss: 4.794 batch_cost: 0.12932 sec reader_cost: 0.00008 sec ips: 7.73275 images/s eta: 3:31:06
[01/18 14:33:11] ppgan.engine.trainer INFO: Iter: 2100/100000 lr: 1.966e-04 loss: 6.932 batch_cost: 0.13047 sec reader_cost: 0.00008 sec ips: 7.66441 images/s eta: 3:32:53
[01/18 14:33:18] ppgan.engine.trainer INFO: Iter: 2150/100000 lr: 1.964e-04 loss: 11.207 batch_cost: 0.13163 sec reader_cost: 0.00009 sec ips: 7.59711 images/s eta: 3:34:39
[01/18 14:33:24] ppgan.engine.trainer INFO: Iter: 2200/100000 lr: 1.962e-04 loss: 11.202 batch_cost: 0.13088 sec reader_cost: 0.00008 sec ips: 7.64079 images/s eta: 3:33:19
[01/18 14:33:31] ppgan.engine.trainer INFO: Iter: 2250/100000 lr: 1.961e-04 loss: 11.765 batch_cost: 0.13034 sec reader_cost: 0.00008 sec ips: 7.67220 images/s eta: 3:32:20
[01/18 14:33:37] ppgan.engine.trainer INFO: Iter: 2300/100000 lr: 1.959e-04 loss: 12.314 batch_cost: 0.13326 sec reader_cost: 0.00009 sec ips: 7.50428 images/s eta: 3:36:59
[01/18 14:33:44] ppgan.engine.trainer INFO: Iter: 2350/100000 lr: 1.957e-04 loss: 3.787 batch_cost: 0.13138 sec reader_cost: 0.00009 sec ips: 7.61134 images/s eta: 3:33:49
[01/18 14:33:50] ppgan.engine.trainer INFO: Iter: 2400/100000 lr: 1.955e-04 loss: 11.284 batch_cost: 0.12889 sec reader_cost: 0.00008 sec ips: 7.75870 images/s eta: 3:29:39
[01/18 14:33:57] ppgan.engine.trainer INFO: Iter: 2450/100000 lr: 1.953e-04 loss: 22.345 batch_cost: 0.12972 sec reader_cost: 0.00008 sec ips: 7.70919 images/s eta: 3:30:53
[01/18 14:34:03] ppgan.engine.trainer INFO: Iter: 2500/100000 lr: 1.951e-04 loss: 16.613 batch_cost: 0.12862 sec reader_cost: 0.00008 sec ips: 7.77481 images/s eta: 3:29:00
Exception in thread Thread-2:
Traceback (most recent call last):
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/threading.py", line 926, in _bootstrap_inner
    self.run()
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/threading.py", line 870, in run
    self._target(*self._args, **self._kwargs)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/dataloader_iter.py", line 213, in _thread_loop
    self._thread_done_event)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/fetcher.py", line 121, in fetch
    data.append(self.dataset[idx])
  File "/home/aistudio/work/PaddleGAN/ppgan/datasets/mpr_dataset.py", line 105, in __getitem__
    inp_img = to_tensor(inp_img)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/vision/transforms/functional.py", line 82, in to_tensor
    return F_pil.to_tensor(pic, data_format)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/vision/transforms/functional_pil.py", line 88, in to_tensor
    img = paddle.cast(img, np.float32) / 255.
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/layers/tensor.py", line 249, in cast
    out = _C_ops.cast(x, 'in_dtype', x.dtype, 'out_dtype', dtype)
SystemError: (Fatal) Operator cast raises an paddle::memory::allocation::BadAlloc exception.
The exception content is
:ResourceExhaustedError: 

Out of memory error on GPU 0. Cannot allocate 3.000244MB memory on GPU 0, 31.716675GB memory has been allocated and available memory is only 2.625000MB.

Please check whether there is any other process using GPU 0.
1. If yes, please stop them, or start PaddlePaddle on another GPU.
2. If no, please decrease the batch size of your model. 

 (at /paddle/paddle/fluid/memory/allocation/cuda_allocator.cc:79)
. (at /paddle/paddle/fluid/imperative/tracer.cc:221)


Traceback (most recent call last):
  File "tools/main.py", line 56, in <module>
    main(args, cfg)
  File "tools/main.py", line 46, in main
    trainer.train()
  File "/home/aistudio/work/PaddleGAN/ppgan/engine/trainer.py", line 191, in train
    self.model.train_iter(self.optimizers)
  File "/home/aistudio/work/PaddleGAN/ppgan/models/mpr_model.py", line 59, in train_iter
    restored = self.nets['generator'](self.lq)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/home/aistudio/work/PaddleGAN/ppgan/models/generators/mpr.py", line 510, in forward
    x3_cat = self.stage3_orsnet(x3_cat, feat2, res2)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/home/aistudio/work/PaddleGAN/ppgan/models/generators/mpr.py", line 338, in forward
    x = self.orb3(x)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/home/aistudio/work/PaddleGAN/ppgan/models/generators/mpr.py", line 276, in forward
    res = self.body(x)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/container.py", line 98, in forward
    input = layer(input)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/home/aistudio/work/PaddleGAN/ppgan/models/generators/mpr.py", line 65, in forward
    res = self.body(x)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/container.py", line 98, in forward
    input = layer(input)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in __call__
    outputs = self.forward(*inputs, **kwargs)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/layer/conv.py", line 677, in forward
    use_cudnn=self._use_cudnn)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/functional/conv.py", line 123, in _conv_nd
    pre_bias = getattr(_C_ops, op_type)(x, weight, *attrs)
SystemError: (Fatal) Operator conv2d raises an paddle::memory::allocation::BadAlloc exception.
The exception content is
:ResourceExhaustedError: 

Out of memory error on GPU 0. Cannot allocate 334.750000kB memory on GPU 0, 31.716675GB memory has been allocated and available memory is only 2.625000MB.

Please check whether there is any other process using GPU 0.
1. If yes, please stop them, or start PaddlePaddle on another GPU.
2. If no, please decrease the batch size of your model. 

 (at /paddle/paddle/fluid/memory/allocation/cuda_allocator.cc:79)
. (at /paddle/paddle/fluid/imperative/tracer.cc:221)

按照官方教程一步步跑示例

Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations if data.dtype == np.object: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:02<00:00, 2.75s/it] 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:02<00:00, 2.75s/it]

C++ Traceback (most recent call last):

0 paddle::framework::SignalHandle(char const*, int) 1 paddle::platform::GetCurrentTraceBackStringabi:cxx11

Error Message Summary:

FatalError: Segmentation fault is detected by the operating system. [TimeInfo: *** Aborted at 1620281429 (unix time) try "date -d @1620281429" if you are using GNU date ***] [SignalInfo: *** SIGSEGV (@0x0) received by PID 32030 (TID 0x7effcb6f4740) from PID 0 ***]

code is like: python applications/tools/animeganv2.py --input_image path/to/gakki.png --output_path path/to/animeGAN errors are:

--------------------------------------
C++ Traceback (most recent call last):
--------------------------------------
0   paddle::imperative::Tracer::TraceOp(std::string const&, paddle::imperative::NameVarBaseMap const&, paddle::imperative::NameVarBaseMap const&, paddle::framework::AttributeMap, std::map<std::string, std::string, std::less<std::string >, std::allocator<std::pair<std::string const, std::string > > > const&)
1   paddle::imperative::Tracer::TraceOp(std::string const&, paddle::imperative::NameVarBaseMap const&, paddle::imperative::NameVarBaseMap const&, paddle::framework::AttributeMap, paddle::platform::Place const&, bool, std::map<std::string, std::string, std::less<std::string >, std::allocator<std::pair<std::string const, std::string > > > const&)
2   paddle::imperative::PreparedOp::Run(paddle::imperative::NameVarBaseMap const&, paddle::imperative::NameVarBaseMap const&, paddle::framework::AttributeMap const&)
3   std::_Function_handler<void (paddle::framework::ExecutionContext const&), paddle::framework::OpKernelRegistrarFunctor<paddle::platform::CPUPlace, false, 0ul, paddle::operators::GemmConvKernel<paddle::platform::CPUDeviceContext, float>, paddle::operators::GemmConvKernel<paddle::platform::CPUDeviceContext, double> >::operator()(char const*, char const*, int) const::{lambda(paddle::framework::ExecutionContext const&)#1}>::_M_invoke(std::_Any_data const&, paddle::framework::ExecutionContext const&)
4   paddle::operators::GemmConvKernel<paddle::platform::CPUDeviceContext, float>::Compute(paddle::framework::ExecutionContext const&) const
5   paddle::operators::math::Im2ColFunctor<(paddle::operators::math::ColFormat)0, paddle::platform::CPUDeviceContext, float>::operator()(paddle::platform::CPUDeviceContext const&, paddle::framework::Tensor const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, paddle::framework::Tensor*, paddle::framework::DataLayout)
6   paddle::framework::SignalHandle(char const*, int)
7   paddle::platform::GetCurrentTraceBackString[abi:cxx11]()

----------------------
Error Message Summary:
----------------------
FatalError: `Segmentation fault` is detected by the operating system.
  [TimeInfo: *** Aborted at 1615879561 (unix time) try "date -d @1615879561" if you are using GNU date ***]
  [SignalInfo: *** SIGSEGV (@0x7f56573e42c0) received by PID 11689 (TID 0x7f60fec05740) from PID 1463698112 ***]

Segmentation fault (core dumped)

my environment: ubuntu 18.04,python 3.8.5,paddlepaddle-gpu 2.0.0

docs/zh_CN/tutorials/pixel2style2pixel.md里的参数配置:

cd applications/
python -u tools/styleganv2.py \
       --input_image <替换为输入的图像路径> \
       --output_path <替换为生成图片存放的文件夹> \
       --weight_path <替换为你的预训练模型路径> \
       --model_type ffhq-inversion \
       --seed 233 \
       --size 1024 \
       --style_dim 512 \
       --n_mlp 8 \
       --channel_multiplier 2 \
       --cpu

可实际所调用的tools/styleganv2.py里并无参数input_image?

(zh version after en | 英文后面是中文)

Why? | 为什么？

Locking librosa strictly to v0.8.1 makes PaddleGAN more difficult to be used with other packages.

把 librosa 版本严格锁在 v0.8.1 导致 PaddleGAN 更难和其它包一起使用。

Is it safe? | 这安全吗？

We only use librosa to read/write wav! Search librosa in the repo.

(You can check the change log if you want)

~~我们用 librosa 只是为了读写 wav！~~在仓库中搜索 librosa。

（也可再检查一下更新日志）

History of the librosa version | librosa 版本的历史

1. Add librosa v0.7.0: add requirement for wav2lip (#147) @a8ba73b
2. Free librosa version: fix numba install in py39 (#427) @5a639c0
3. Lock librosa to v0.8.1: update librosa version (#504) @0235853

1. 添加 librosa v0.7.0: 添加 wav2lip 的依赖 (#147) @a8ba73b
2. 放松 librosa 版本: fix numba install in py39 (#427) @5a639c0
3. 锁定 librosa 到 v0.8.1: 更新 librosa 版本 (#504) @0235853

问题描述： 使用PaddleGAN中mprnet_deblurring.yaml进行训练，选用了GoPro数据集，未修改其余参数。 使用环境： AI Studio在线平台、PaddlePaddle2.2.0、Python3.7 代码如下：

从github上克隆PaddleGAN代码 !git clone https://gitee.com/paddlepaddle/PaddleGAN 安装所需安装包 %cd PaddleGAN/ !pip install -r requirements.txt !pip install imageio-ffmpeg %cd applications/ 单卡训练 !python -u PaddleGAN/tools/main.py --config-file work/mprnet_deblurring.yaml

报错如下：

[11/18 15:19:07] ppgan INFO: Configs: {'total_iters': 100000, 'output_dir': 'output_dir/mprnet_deblurring-2021-11-18-15-19', 'model': {'name': 'MPRModel', 'generator': {'name': 'MPRNet'}, 'char_criterion': {'name': 'CharbonnierLoss'}, 'edge_criterion': {'name': 'EdgeLoss'}}, 'dataset': {'train': {'name': 'MPRTrain', 'rgb_dir': 'GoPro', 'num_workers': 16, 'batch_size': 4, 'img_options': {'patch_size': 256}}, 'test': {'name': 'MPRTrain', 'rgb_dir': 'GoPro', 'num_workers': 16, 'batch_size': 4, 'img_options': {'patch_size': 256}}}, 'lr_scheduler': {'name': 'CosineAnnealingRestartLR', 'learning_rate': 0.0002, 'periods': [25000, 25000, 25000, 25000], 'restart_weights': [1, 1, 1, 1], 'eta_min': 1e-06}, 'validate': {'interval': 10, 'save_img': False, 'metrics': {'psnr': {'name': 'PSNR', 'crop_border': 4, 'test_y_channel': True}, 'ssim': {'name': 'SSIM', 'crop_border': 4, 'test_y_channel': True}}}, 'optimizer': {'name': 'Adam', 'net_names': ['generator'], 'beta1': 0.9, 'beta2': 0.999, 'epsilon': 1e-08}, 'log_config': {'interval': 10, 'visiual_interval': 5000}, 'snapshot_config': {'interval': 5000}, 'is_train': True, 'profiler_options': None, 'timestamp': '-2021-11-18-15-19'} W1118 15:19:07.026855 714 device_context.cc:404] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 10.1, Runtime API Version: 10.1 W1118 15:19:07.031368 714 device_context.cc:422] device: 0, cuDNN Version: 7.6. /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/tensor/creation.py:125: DeprecationWarning: np.object is a deprecated alias for the builtin object. To silence this warning, use object by itself. Doing this will not modify any behavior and is safe. Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations if data.dtype == np.object: Traceback (most recent call last): File "PaddleGAN/tools/main.py", line 56, in main(args, cfg) File "PaddleGAN/tools/main.py", line 46, in main trainer.train() File "/home/aistudio/PaddleGAN/ppgan/engine/trainer.py", line 190, in train self.model.train_iter(self.optimizers) File "/home/aistudio/PaddleGAN/ppgan/models/mpr_model.py", line 65, in train_iter loss_edge.append(self.edge_criterion(restored[i], self.target)) File "/home/aistudio/PaddleGAN/ppgan/models/criterions/pixel_loss.py", line 270, in call loss = self.loss(self.laplacian_kernel(x), self.laplacian_kernel(y)) File "/home/aistudio/PaddleGAN/ppgan/models/criterions/pixel_loss.py", line 264, in laplacian_kernel new_filter[:,:,::2,::2] = down*4 # upsample ValueError: (InvalidArgument) Leaf Tensor (dygraph_tmp_1341) that doesn't stop gradient can't use inplace strategy. [Hint: Expected self->IsLeaf() && !self->OverridedStopGradient() == false, but received self->IsLeaf() && !self->OverridedStopGradient():1 != false:0.] (at /paddle/paddle/fluid/pybind/imperative.cc:718)

thank you for great project! I have issue with installing ppgan package with python3.10, because there is no such version for numba package on pypi. Could you please relax it to make it compatibe with new python versions?

Traceback (most recent call last): File "tools/main.py", line 56, in main(args, cfg) File "tools/main.py", line 46, in main Exception in thread Thread-1: Traceback (most recent call last): File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/threading.py", line 926, in _bootstrap_inner self.run() File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/threading.py", line 870, in run self._target(*self._args, **self._kwargs) File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/dataloader_iter.py", line 534, in _thread_loop batch = self._get_data() File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/dataloader_iter.py", line 671, in _get_data batch.reraise() File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/worker.py", line 169, in reraise raise self.exc_type(msg) ValueError: DataLoader worker(0) caught ValueError with message: Traceback (most recent call last): File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/worker.py", line 336, in _worker_loop batch = fetcher.fetch(indices) File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dataloader/fetcher.py", line 121, in fetch data.append(self.dataset[idx]) File "/home/aistudio/PaddleGAN-develop/ppgan/datasets/nafnet_dataset.py", line 64, in getitem padding_mode='reflect') File "<array_function internals>", line 6, in pad File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/numpy/lib/arraypad.py", line 746, in pad pad_width = _as_pairs(pad_width, array.ndim, as_index=True) File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/numpy/lib/arraypad.py", line 521, in _as_pairs return np.broadcast_to(x, (ndim, 2)).tolist() File "<array_function internals>", line 6, in broadcast_to File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/numpy/lib/stride_tricks.py", line 180, in broadcast_to return _broadcast_to(array, shape, subok=subok, readonly=True) File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/numpy/lib/stride_tricks.py", line 125, in _broadcast_to op_flags=['readonly'], itershape=shape, order='C') ValueError: operands could not be broadcast together with remapped shapes [original->remapped]: (4,) and requested shape (3,2)

800张训练集，batch_size 8，patch_size 100 时不会出现 当 patch_size 设置为更大的数时会出现

疑似错误代码：

        # Reflect Pad in case image is smaller than patch_size
        if padw != 0 or padh != 0:
            inp_img = np.pad(inp_img, (0, 0, padw, padh),
                             padding_mode='reflect')
            tar_img = np.pad(tar_img, (0, 0, padw, padh),
                             padding_mode='reflect')

在以下文档中所有的视频超分模型如何导出呀？

https://github.com/PaddlePaddle/PaddleGAN/blob/develop/docs/zh_CN/tutorials/video_super_resolution.md

环境：cpu环境 版本：paddlegan develop 导出命令：

python tools/export_model.py -c configs/msvsr_l_reds.yaml --inputs_size="1,2,3,180,320" --model_name model --load ~/Downloads/PP-MSVSR-L_reds_x4.pdparams

问题：模型导出除了按照教程中的pp-msvsr导出模型成功，其他pp-msvsr-l 等等 其它模型导出错误，提示

[11/07 10:08:28] ppgan INFO: Found /Users/aichao/.cache/ppgan/modified_spynet.pdparams
Traceback (most recent call last):
  File "./ppgan/utils/config.py", line 23, in __getattr__
    return self[key]
KeyError: 'export_model'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "tools/export_model.py", line 88, in <module>
    main(args, cfg)
  File "tools/export_model.py", line 81, in main
    model.export_model(cfg.export_model, args.output_dir, inputs_size,
  File "./ppgan/utils/config.py", line 25, in __getattr__
    raise AttributeError(key)
AttributeError: export_model