paddle-gan在windows运行StarGAN V2算法报错

[xiaoxiaotom]

(paddlegan) PS F:\PaddleGAN> python -u tools/main.py --config-file configs/starganv2_celeba_hq.yaml
[11/21 11:29:27] ppgan INFO: Configs:
dataset:
test:
batch_size: 16
dataroot: data/stargan-v2/celeba_hq/val/
is_train: false
name: StarGANv2Dataset
num_workers: 8
preprocess:
- key: src
name: LoadImageFromFile
- key: ref
name: LoadImageFromFile
- input_keys:
- src
- ref
name: Transforms
pipeline:
- interpolation: bicubic
keys:
- image
- image
name: Resize
size:
- 256
- 256
- keys:
- image
- image
name: Transpose
- keys:
- image
- image
mean:
- 127.5
- 127.5
- 127.5
name: Normalize
std:
- 127.5
- 127.5
- 127.5
test_count: 16
train:
batch_size: 4
dataroot: data/stargan-v2/celeba_hq/train/
is_train: true
name: StarGANv2Dataset
num_workers: 8
preprocess:
- key: src
name: LoadImageFromFile
- key: ref
name: LoadImageFromFile
- key: ref2
name: LoadImageFromFile
- input_keys:
- src
- ref
- ref2
name: Transforms
pipeline:
- interpolation: bilinear
keys:
- image
- image
- image
name: RandomResizedCropProb
prob: 0.9
ratio:
- 0.9
- 1.1
scale:
- 0.8
- 1.0
size:
- 256
- 256
- interpolation: bilinear
keys:
- image
- image
- image
name: Resize
size:
- 256
- 256
- keys:
- image
- image
- image
name: RandomHorizontalFlip
prob: 0.5
- keys:
- image
- image
- image
name: Transpose
- keys:
- image
- image
- image
mean:
- 127.5
- 127.5
- 127.5
name: Normalize
std:
- 127.5
- 127.5
- 127.5
epochs: 200
is_train: true
log_config:
interval: 100
visiual_interval: 3000
lr_scheduler:
decay_epochs: 100
iters_per_epoch: 365
learning_rate: 0.0001
name: LinearDecay
start_epoch: 100
model:
discriminator:
img_size: 256
name: StarGANv2Discriminator
num_domains: 2
fan:
fname_pretrained: null
name: FAN
generator:
img_size: 256
name: StarGANv2Generator
style_dim: 64
w_hpf: 1
lambda_cyc: 1
lambda_ds: 1
lambda_sty: 1
latent_dim: 16
mapping:
latent_dim: 16
name: StarGANv2Mapping
num_domains: 2
style_dim: 64
name: StarGANv2Model
style:
img_size: 256
name: StarGANv2Style
num_domains: 2
style_dim: 64
optimizer:
discriminator:
beta1: 0.0
beta2: 0.99
name: Adam
net_names:
- discriminator
weight_decay: 0.0001
generator:
beta1: 0.0
beta2: 0.99
name: Adam
net_names:
- generator
weight_decay: 0.0001
mapping_network:
beta1: 0.0
beta2: 0.99
name: Adam
net_names:
- mapping_network
weight_decay: 0.0001
main(args, cfg)
File "F:\PaddleGAN\tools\main.py", line 46, in main
trainer.train()
File "F:\PaddleGAN\ppgan\engine\trainer.py", line 234, in train
self.model.train_iter(self.optimizers) # norm train
File "F:\PaddleGAN\ppgan\models\starganv2_model.py", line 279, in train_iter
d_loss, d_losses_latent = compute_d_loss(self.nets,
File "F:\PaddleGAN\ppgan\models\starganv2_model.py", line 64, in compute_d_loss
s_trg = nets['mapping_network'](z_trg, y_trg)
File "D:\ProgramData\miniconda3\envs\paddlegan\lib\site-packages\paddle\nn\layer\layers.py", line 1426, in call
return self.forward(*inputs, **kwargs)
File "F:\PaddleGAN\ppgan\models\generators\generator_starganv2.py", line 245, in forward
out[idx[i].numpy().astype(np.int_).tolist()[0],
TypeError: 'int' object is not subscriptable
感觉是numpy的版本问题，请问应该选择什么版本