Training a GAN - "Your generator is empty" error

Deep Learning
1

Hi all,

I’m trying to train a GAN to generate synthetic images. This is my first foray into GANs and I may have taken Jeremy’s “just try some code” advice too far, but I can’t work out the issue here.

Here’s my code:

from fastbook import *
from google.colab import drive
import requests
import json
from fastai.vision import *
from fastai.vision.gan import *

drive.mount('/content/gdrive', force_remount=True)
root_dir = "/content/gdrive/My Drive/"
base_dir = root_dir + 'Colab/GAN/'

%reload_ext autoreload
%autoreload 2
%matplotlib inline

# I've tried making batch size smaller and smaller, no effect
bs = 2
size = 32

dblock = DataBlock(blocks = (TransformBlock, ImageBlock),
                   get_x = generate_noise,
                   get_items = get_image_files,
                   splitter = IndexSplitter([]),
                   item_tfms=Resize(size, method=ResizeMethod.Crop), 
                   batch_tfms = [*aug_transforms(), Normalize.from_stats(torch.tensor([0.5,0.5,0.5]), torch.tensor([0.5,0.5,0.5]))])

dls = dblock.dataloaders(base_dir + 'Images/', bs = bs)

generator = basic_generator(out_size=size, n_channels=3, n_extra_layers=1)
critic    = basic_critic(in_size=size, n_channels=3, n_extra_layers=1)

learn = GANLearner.wgan(dls, generator, critic, switch_eval=False,
                        opt_func = Adam, wd=0.)

learn.fit(3,2e-2)

I have 911 images. The fit progresses with the following warnings:

/usr/local/lib/python3.7/dist-packages/fastai/callback/core.py:51: UserWarning: You are shadowing an attribute (generator) that exists in the learner. Use self.learn.generator to avoid this warn(f"You are shadowing an attribute ({name}) that exists in the learner. Use self.learn.{name} to avoid this") /usr/local/lib/python3.7/dist-packages/fastai/callback/core.py:51: UserWarning: You are shadowing an attribute (critic) that exists in the learner. Use self.learn.critic to avoid this warn(f"You are shadowing an attribute ({name}) that exists in the learner. Use self.learn.{name} to avoid this") /usr/local/lib/python3.7/dist-packages/fastai/callback/core.py:51: UserWarning: You are shadowing an attribute (gen_mode) that exists in the learner. Use self.learn.gen_mode to avoid this warn(f"You are shadowing an attribute ({name}) that exists in the learner. Use self.learn.{name} to avoid this")

epoch train_loss valid_loss gen_loss crit_loss time
0 -0.462891 None None None 05:35
1 -0.503794 None None None 05:31
2 -0.275025 None None None 05:35

/usr/local/lib/python3.7/dist-packages/PIL/Image.py:2800: DecompressionBombWarning: Image size (99996755 pixels) exceeds limit of 89478485 pixels, could be decompression bomb DOS attack. DecompressionBombWarning, /usr/local/lib/python3.7/dist-packages/fastprogress/fastprogress.py:74: UserWarning: Your generator is empty. warn(“Your generator is empty.”) /usr/local/lib/python3.7/dist-packages/PIL/Image.py:2800: DecompressionBombWarning: Image size (99996755 pixels) exceeds limit of 89478485 pixels, could be decompression bomb DOS attack. DecompressionBombWarning, /usr/local/lib/python3.7/dist-packages/PIL/Image.py:2800: DecompressionBombWarning: Image size (99996755 pixels) exceeds limit of 89478485 pixels, could be decompression bomb DOS attack. DecompressionBombWarning,

Can anyone spot the problem? If I try to view results I get an error as well:

learn.gan_trainer.switch(gen_mode=True)
learn.show_results(rows=16, figsize=(8,8))

/usr/local/lib/python3.7/dist-packages/fastai/callback/core.py:51: UserWarning: You are shadowing an attribute (gen_mode) that exists in the learner. Use self.learn.gen_mode to avoid this
warn(f"You are shadowing an attribute ({name}) that exists in the learner. Use self.learn.{name} to avoid this")

ValueError Traceback (most recent call last)
in ()
1 learn.gan_trainer.switch(gen_mode=True)
----> 2 learn.show_results(rows=16, figsize=(8,8))

1 frames
/usr/local/lib/python3.7/dist-packages/fastai/data/load.py in one_batch(self)
145 def to(self, device): self.device = device
146 def one_batch(self):
→ 147 if self.n is not None and len(self)==0: raise ValueError(f’This DataLoader does not contain any batches’)
148 with self.fake_l.no_multiproc(): res = first(self)
149 if hasattr(self, ‘it’): delattr(self, ‘it’)

ValueError: This DataLoader does not contain any batches

2

Hi Ponty

Try Example for GANLearner / GANItemList?

Regards Conwyn

3

Thanks Conwyn.

Unfortunately GANItemList seems to not exist anymore. This makes things a bit more difficult as all the existing examples use that approach.

4

Can you confirm your DataLoader looks right?

Try dls.show_batch(max_n=16)

5

Hi @ilovescience, yep when I run that command I see a grid of my input images (either an amount equal to bs, which I set at the top, or 16, whichever is smaller). The input images appear distorted due to aug_transforms (if I remove that from the dblock definition, my images appear undistorted).

Furthermore, dls.n gives me 911.