def test_dissection():
verbose_progress(True)
from torchvision.models import alexnet
from torchvision import transforms
model = InstrumentedModel(alexnet(pretrained=True))
model.eval()
# Load an alexnet
model.retain_layers([('features.0', 'conv1'), ('features.3', 'conv2'),
('features.6', 'conv3'), ('features.8', 'conv4'),
('features.10', 'conv5')])
# load broden dataset
bds = BrodenDataset('dataset/broden',
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(IMAGE_MEAN, IMAGE_STDEV)
]),
size=100)
# run dissect
dissect('dissect/test', model, bds, examples_per_unit=10)
def main():
# Training settings
def strpair(arg):
p = tuple(arg.split(':'))
if len(p) == 1:
p = p + p
return p
def intpair(arg):
p = arg.split(',')
if len(p) == 1:
p = p + p
return tuple(int(v) for v in p)
parser = argparse.ArgumentParser(
description='Net dissect utility',
prog='python -m netdissect',
epilog=textwrap.dedent(help_epilog),
formatter_class=argparse.RawDescriptionHelpFormatter)
parser.add_argument('--model',
type=str,
default=None,
help='constructor for the model to test')
parser.add_argument('--pthfile',
type=str,
default=None,
help='filename of .pth file for the model')
parser.add_argument('--outdir',
type=str,
default='dissect',
help='directory for dissection output')
parser.add_argument('--layers',
type=strpair,
nargs='+',
help='space-separated list of layer names to dissect' +
', in the form layername[:reportedname]')
parser.add_argument('--segments',
type=str,
default='dataset/broden',
help='directory containing segmentation dataset')
parser.add_argument('--download',
action='store_true',
default=False,
help='downloads Broden dataset if needed')
parser.add_argument('--imgsize',
type=intpair,
default=(227, 227),
help='input image size to use')
parser.add_argument('--netname',
type=str,
default=None,
help='name for network in generated reports')
parser.add_argument('--meta',
type=str,
nargs='+',
help='json files of metadata to add to report')
parser.add_argument('--examples',
type=int,
default=20,
help='number of image examples per unit')
parser.add_argument('--size',
type=int,
default=10000,
help='dataset subset size to use')
parser.add_argument('--broden_version',
type=int,
default=1,
help='broden version number')
parser.add_argument('--batch_size',
type=int,
default=100,
help='batch size for forward pass')
parser.add_argument('--num_workers',
type=int,
default=24,
help='number of DataLoader workers')
parser.add_argument('--quantile_threshold',
type=float,
default=None,
help='quantile to use for masks')
parser.add_argument('--no-labels',
action='store_true',
default=False,
help='disables labeling of units')
parser.add_argument('--ablation',
action='store_true',
default=False,
help='enables single unit ablation of units')
parser.add_argument('--iqr',
action='store_true',
default=False,
help='enables iqr calculation')
parser.add_argument('--maxiou',
action='store_true',
default=False,
help='enables maxiou calculation')
parser.add_argument('--covariance',
action='store_true',
default=False,
help='enables covariance calculation')
parser.add_argument('--no-images',
action='store_true',
default=False,
help='disables generation of unit images')
parser.add_argument('--single-images',
action='store_true',
default=False,
help='generates single images also')
parser.add_argument('--no-report',
action='store_true',
default=False,
help='disables generation report summary')
parser.add_argument('--no-cuda',
action='store_true',
default=False,
help='disables CUDA usage')
parser.add_argument('--gan',
type=str,
default=None,
help='netdissect.GanImageSegmenter() to probe a GAN')
parser.add_argument('--perturbation',
default=None,
help='filename of perturbation attack to apply')
parser.add_argument('--add_scale_offset',
action='store_true',
default=None,
help='offsets masks according to stride and padding')
parser.add_argument('--quiet',
action='store_true',
default=False,
help='silences console output')
if len(sys.argv) == 1:
parser.print_usage(sys.stderr)
sys.exit(1)
args = parser.parse_args()
args.images = not args.no_images
args.report = not args.no_report
args.labels = not args.no_labels
# Set up console output
verbose_progress(not args.quiet)
# Speed up pytorch
torch.backends.cudnn.benchmark = True
# Special case: download flag without model to test.
if args.model is None and args.download:
from netdissect.broden import ensure_broden_downloaded
for resolution in [224, 227, 384]:
ensure_broden_downloaded(args.segments, resolution,
args.broden_version)
sys.exit(0)
# Help if broden is not present
if not os.path.isdir(args.segments):
print_progress('Segmentation dataset not found at %s.' % args.segments)
print_progress('Specify dataset directory using --segments [DIR]')
print_progress('To download Broden, run: netdissect --download')
sys.exit(1)
# Default threshold
if args.quantile_threshold is None:
if args.gan:
args.quantile_threshold = 0.01
else:
args.quantile_threshold = 0.005
# Construct the network
if args.model is None:
print_progress('No model specified')
sys.exit(1)
# Set up CUDA
args.cuda = not args.no_cuda and torch.cuda.is_available()
if args.cuda:
torch.backends.cudnn.benchmark = True
model = autoimport_eval(args.model)
# Unwrap any DataParallel-wrapped model
if isinstance(model, torch.nn.DataParallel):
model = next(model.children())
# Default add_scale_offset only for AlexNet-looking models.
if args.add_scale_offset is None and not args.gan:
args.add_scale_offset = ('Alex' in model.__class__.__name__)
# Load its state dict
meta = {}
if args.pthfile is None:
print_progress('Dissecting model without pth file.')
else:
data = torch.load(args.pthfile)
if 'state_dict' in data:
meta = {}
for key in data:
if isinstance(data[key], numbers.Number):
meta[key] = data[key]
data = data['state_dict']
model.load_state_dict(data)
# Update any metadata from files, if any
if args.meta:
for mfilename in args.meta:
with open(mfilename) as f:
meta.update(json.load(f))
# Instrument it and prepare it for eval
if not args.layers:
# Skip wrappers with only one named modele
container = model
prefix = ''
while len(list(container.named_children())) == 1:
name, container = next(container.named_children())
prefix += name + '.'
# Default to all nontrivial top-level layers except last.
args.layers = [
prefix + name for name, module in container.named_children()
if type(module).__module__ not in [
# Skip ReLU and other activations.
'torch.nn.modules.activation',
# Skip pooling layers.
'torch.nn.modules.pooling'
]
][:-1]
print_progress('Defaulting to layers: %s' % ' '.join(args.layers))
retain_layers(model, args.layers, args.add_scale_offset)
if args.gan:
ablate_layers(model, args.layers)
model.eval()
if args.cuda:
model.cuda()
# Set up the output directory, verify write access
if args.outdir is None:
args.outdir = os.path.join('dissect', type(model).__name__)
print_progress('Writing output into %s.' % args.outdir)
os.makedirs(args.outdir, exist_ok=True)
train_dataset = None
if not args.gan:
# Load dataset for ordinary case.
# Load perturbation
perturbation = numpy.load(
args.perturbation) if args.perturbation else None
# Load broden dataset
dataset = try_to_load_broden(args.segments, args.imgsize,
args.broden_version, perturbation,
args.download, args.size)
if dataset is None:
ds = try_to_load_multiseg(args.segments, args.imgsize,
perturbation, args.size)
if dataset is None:
print_progress('No segmentation dataset found in %s' %
args.segements)
print_progress('use --download to download Broden.')
sys.exit(1)
recovery = ReverseNormalize(IMAGE_MEAN, IMAGE_STDEV)
else:
# Examine first conv in model to determine input feature size.
first_layer = [
c for c in model.modules()
if isinstance(c, (torch.nn.Conv2d, torch.nn.ConvTranspose2d,
torch.nn.Linear))
][0]
# 4d input if convolutional, 2d input if first layer is linear.
if isinstance(first_layer,
(torch.nn.Conv2d, torch.nn.ConvTranspose2d)):
sample = standard_z_sample(args.size,
first_layer.in_channels)[:, :, None,
None]
train_sample = standard_z_sample(args.size,
first_layer.in_channels,
seed=2)[:, :, None, None]
else:
sample = standard_z_sample(args.size, first_layer.in_features)
train_sample = standard_z_sample(args.size,
first_layer.in_features,
seed=2)
dataset = TensorDataset(sample)
train_dataset = TensorDataset(train_sample)
recovery = autoimport_eval(args.gan)
# Run dissect
dissect(args.outdir,
model,
dataset,
train_dataset=train_dataset,
recover_image=recovery,
examples_per_unit=args.examples,
netname=args.netname,
quantile_threshold=args.quantile_threshold,
meta=meta,
make_images=args.images or args.single_images,
make_labels=args.labels,
make_ablation=args.ablation,
make_iqr=args.iqr,
make_maxiou=args.maxiou,
make_covariance=args.covariance,
make_report=args.report,
make_row_images=args.images,
make_single_images=args.single_images,
batch_size=args.batch_size,
num_workers=args.num_workers,
settings=vars(args))
def optimize_neurons(self):
# Set up console output
verbose_progress(True)
gan_model = self.generator.model
annotate_model_shapes(gan_model, gen=True)
outdir = os.path.join(
self.args.results, 'dissect',
self.args.name_checkpoint + '_' + str(time.time()))
os.makedirs(outdir, exist_ok=True)
size = 1000
sample = z_sample_for_model(gan_model, size)
train_sample = z_sample_for_model(gan_model, size, seed=2)
dataset = TensorDataset(sample)
train_dataset = TensorDataset(train_sample)
self.cluster_segmenter = ClusterSegmenter(self.model, self.clusters,
self.mean_clust,
self.std_clust)
segrunner = GeneratorSegRunner(self.cluster_segmenter)
netname = outdir
# Run dissect
with torch.no_grad():
dissect(
outdir,
gan_model,
dataset,
train_dataset=train_dataset,
segrunner=segrunner,
examples_per_unit=20,
netname=netname,
quantile_threshold='iqr',
meta=None,
make_images=False, # True,
make_labels=True,
make_maxiou=False,
make_covariance=False,
make_report=True,
make_row_images=True,
make_single_images=True,
batch_size=8,
num_workers=8,
rank_all_labels=True)
sample_ablate = z_sample_for_model(gan_model, 16)
dataset_ablate = TensorDataset(sample_ablate)
data_loader = torch.utils.data.DataLoader(dataset_ablate,
batch_size=8,
shuffle=False,
num_workers=8,
pin_memory=True,
sampler=None)
with open(os.path.join(outdir, 'dissect.json')) as f:
data = EasyDict(json.load(f))
dissect_layer = {lrec.layer: lrec for lrec in data.layers}
self.layers_units = {
'layer2': [],
'layer3': [],
'layer4': [],
'layer5': [],
'layer6': [],
}
noise_units = np.array([35, 221, 496, 280])
for i in range(2, len(self.clusters) + 2):
print('Cluster', i)
rank_name = 'c_{0}-iou'.format(i)
for l in range(len(self.layer_list_all)):
ranking = next(
r
for r in dissect_layer[self.layer_list_all[l]].rankings
if r.name == rank_name)
unit_list = np.array(range(512))
unit_list[noise_units] = 0
ordering = np.argsort(ranking.score)
units_list = unit_list[ordering]
self.layers_units[self.layer_list_all[l]].append(
units_list)
# Mark the directory so that it's not done again.
mark_job_done(outdir)
def main():
# Training settings
def strpair(arg):
p = tuple(arg.split(':'))
if len(p) == 1:
p = p + p
return p
def intpair(arg):
p = arg.split(',')
if len(p) == 1:
p = p + p
return tuple(int(v) for v in p)
parser = argparse.ArgumentParser(
description='Net dissect utility',
prog='python -m netdissect',
epilog=textwrap.dedent(help_epilog),
formatter_class=argparse.RawDescriptionHelpFormatter)
parser.add_argument('--model',
type=str,
default=None,
help='constructor for the model to test')
parser.add_argument('--pthfile',
type=str,
default=None,
help='filename of .pth file for the model')
parser.add_argument('--unstrict',
action='store_true',
default=False,
help='ignore unexpected pth parameters')
parser.add_argument('--submodule',
type=str,
default=None,
help='submodule to load from pthfile')
parser.add_argument('--outdir',
type=str,
default='dissect',
help='directory for dissection output')
parser.add_argument('--layers',
type=strpair,
nargs='+',
help='space-separated list of layer names to dissect' +
', in the form layername[:reportedname]')
parser.add_argument('--segments',
type=str,
default='dataset/broden',
help='directory containing segmentation dataset')
parser.add_argument('--segmenter',
type=str,
default=None,
help='constructor for asegmenter class')
parser.add_argument('--download',
action='store_true',
default=False,
help='downloads Broden dataset if needed')
parser.add_argument('--imagedir',
type=str,
default=None,
help='directory containing image-only dataset')
parser.add_argument('--imgsize',
type=intpair,
default=(227, 227),
help='input image size to use')
parser.add_argument('--netname',
type=str,
default=None,
help='name for network in generated reports')
parser.add_argument('--meta',
type=str,
nargs='+',
help='json files of metadata to add to report')
parser.add_argument('--merge',
type=str,
help='json file of unit data to merge in report')
parser.add_argument('--examples',
type=int,
default=20,
help='number of image examples per unit')
parser.add_argument('--size',
type=int,
default=10000,
help='dataset subset size to use')
parser.add_argument('--batch_size',
type=int,
default=100,
help='batch size for forward pass')
parser.add_argument('--num_workers',
type=int,
default=24,
help='number of DataLoader workers')
parser.add_argument('--quantile_threshold',
type=strfloat,
default=None,
choices=[FloatRange(0.0, 1.0), 'iqr'],
help='quantile to use for masks')
parser.add_argument('--no-labels',
action='store_true',
default=False,
help='disables labeling of units')
parser.add_argument('--maxiou',
action='store_true',
default=False,
help='enables maxiou calculation')
parser.add_argument('--covariance',
action='store_true',
default=False,
help='enables covariance calculation')
parser.add_argument('--rank_all_labels',
action='store_true',
default=False,
help='include low-information labels in rankings')
parser.add_argument('--no-images',
action='store_true',
default=False,
help='disables generation of unit images')
parser.add_argument('--no-report',
action='store_true',
default=False,
help='disables generation report summary')
parser.add_argument('--no-cuda',
action='store_true',
default=False,
help='disables CUDA usage')
parser.add_argument('--gen',
action='store_true',
default=False,
help='test a generator model (e.g., a GAN)')
parser.add_argument('--gan',
action='store_true',
default=False,
help='synonym for --gen')
parser.add_argument('--perturbation',
default=None,
help='filename of perturbation attack to apply')
parser.add_argument('--add_scale_offset',
action='store_true',
default=None,
help='offsets masks according to stride and padding')
parser.add_argument('--quiet',
action='store_true',
default=False,
help='silences console output')
if len(sys.argv) == 1:
parser.print_usage(sys.stderr)
sys.exit(1)
args = parser.parse_args()
args.images = not args.no_images
args.report = not args.no_report
args.labels = not args.no_labels
if args.gan:
args.gen = args.gan
# Set up console output
verbose_progress(not args.quiet)
# Exit right away if job is already done or being done.
if args.outdir is not None:
exit_if_job_done(args.outdir)
# Speed up pytorch
torch.backends.cudnn.benchmark = True
# Special case: download flag without model to test.
if args.model is None and args.download:
from netdissect.broden import ensure_broden_downloaded
for resolution in [224, 227, 384]:
ensure_broden_downloaded(args.segments, resolution, 1)
from netdissect.segmenter import ensure_upp_segmenter_downloaded
ensure_upp_segmenter_downloaded('dataset/segmodel')
sys.exit(0)
# Help if broden is not present
if not args.gen and not args.imagedir and not os.path.isdir(args.segments):
print_progress('Segmentation dataset not found at %s.' % args.segments)
print_progress('Specify dataset directory using --segments [DIR]')
print_progress('To download Broden, run: netdissect --download')
sys.exit(1)
# Default segmenter class
if args.gen and args.segmenter is None:
args.segmenter = ("netdissect.segmenter.UnifiedParsingSegmenter(" +
"segsizes=[256], segdiv=None)")
# Default threshold
if args.quantile_threshold is None:
if args.gen:
args.quantile_threshold = 'iqr'
else:
args.quantile_threshold = 0.005
# Set up CUDA
args.cuda = not args.no_cuda and torch.cuda.is_available()
if args.cuda:
torch.backends.cudnn.benchmark = True
# Construct the network with specified layers instrumented
if args.model is None:
print_progress('No model specified')
sys.exit(1)
model = create_instrumented_model(args)
# Update any metadata from files, if any
meta = getattr(model, 'meta', {})
if args.meta:
for mfilename in args.meta:
with open(mfilename) as f:
meta.update(json.load(f))
# Load any merge data from files
mergedata = None
if args.merge:
with open(args.merge) as f:
mergedata = json.load(f)
# Set up the output directory, verify write access
if args.outdir is None:
args.outdir = os.path.join('dissect', type(model).__name__)
exit_if_job_done(args.outdir)
print_progress('Writing output into %s.' % args.outdir)
os.makedirs(args.outdir, exist_ok=True)
train_dataset = None
if not args.gen:
# Load dataset for classifier case.
# Load perturbation
perturbation = numpy.load(
args.perturbation) if args.perturbation else None
segrunner = None
# Load broden dataset
if args.imagedir is not None:
dataset = try_to_load_images(args.imagedir, args.imgsize,
perturbation, args.size)
segrunner = ImageOnlySegRunner(dataset)
else:
dataset = try_to_load_broden(args.segments, args.imgsize, 1,
perturbation, args.download,
args.size)
if dataset is None:
dataset = try_to_load_multiseg(args.segments, args.imgsize,
perturbation, args.size)
if dataset is None:
print_progress('No segmentation dataset found in %s',
args.segments)
print_progress('use --download to download Broden.')
sys.exit(1)
else:
# For segmenter case the dataset is just a random z
#dataset = z_dataset_for_model(model, args.size)
#train_dataset = z_dataset_for_model(model, args.size, seed=2)
dataset = datasets.ImageFolder('dataset/Adissect',
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(
(0.5, 0.5, 0.5),
(0.5, 0.5, 0.5))
]))
train_dataset = dataset
segrunner = GeneratorSegRunner(autoimport_eval(args.segmenter))
torch.no_grad()
# Run dissect
dissect(args.outdir,
model,
dataset,
train_dataset=train_dataset,
segrunner=segrunner,
examples_per_unit=args.examples,
netname=args.netname,
quantile_threshold=args.quantile_threshold,
meta=meta,
merge=mergedata,
make_images=args.images,
make_labels=args.labels,
make_maxiou=args.maxiou,
make_covariance=args.covariance,
make_report=args.report,
make_row_images=args.images,
make_single_images=True,
rank_all_labels=args.rank_all_labels,
batch_size=args.batch_size,
num_workers=args.num_workers,
settings=vars(args))
# Mark the directory so that it's not done again.
mark_job_done(args.outdir)
def main():
# Training settings
def strpair(arg):
p = tuple(arg.split(':'))
if len(p) == 1:
p = p + p
return p
def intpair(arg):
p = arg.split(',')
if len(p) == 1:
p = p + p
return tuple(int(v) for v in p)
parser = argparse.ArgumentParser(
description='Net dissect utility',
prog='python -m netdissect',
epilog=textwrap.dedent(help_epilog),
formatter_class=argparse.RawDescriptionHelpFormatter)
parser.add_argument('--model',
type=str,
default=None,
help='constructor for the model to test')
parser.add_argument('--pthfile',
type=str,
default=None,
help='filename of .pth file for the model')
parser.add_argument('--outdir',
type=str,
default='dissect',
help='directory for dissection output')
parser.add_argument('--layers',
type=strpair,
nargs='+',
help='space-separated list of layer names to dissect' +
', in the form layername[:reportedname]')
parser.add_argument('--segments',
type=str,
default='dataset/broden',
help='directory containing segmentation dataset')
parser.add_argument('--download',
action='store_true',
default=False,
help='downloads Broden dataset if needed')
parser.add_argument('--imgsize',
type=intpair,
default=(227, 227),
help='input image size to use')
parser.add_argument('--netname',
type=str,
default=None,
help='name for network in generated reports')
parser.add_argument('--meta',
type=str,
nargs='+',
help='json files of metadata to add to report')
parser.add_argument('--examples',
type=int,
default=20,
help='number of image examples per unit')
parser.add_argument('--size',
type=int,
default=10000,
help='dataset subset size to use')
parser.add_argument('--broden_version',
type=int,
default=1,
help='broden version number')
parser.add_argument('--batch_size',
type=int,
default=100,
help='batch size for forward pass')
parser.add_argument('--num_workers',
type=int,
default=24,
help='number of DataLoader workers')
parser.add_argument('--no-labels',
action='store_true',
default=False,
help='disables labeling of units')
parser.add_argument('--no-images',
action='store_true',
default=False,
help='disables generation of unit images')
parser.add_argument('--single-images',
action='store_true',
default=False,
help='generates single images also')
parser.add_argument('--no-report',
action='store_true',
default=False,
help='disables generation report summary')
parser.add_argument('--no-cuda',
action='store_true',
default=False,
help='disables CUDA usage')
parser.add_argument('--perturbation',
default=None,
help='filename of perturbation attack to apply')
parser.add_argument('--add_scale_offset',
action='store_true',
default=None,
help='offsets masks according to stride and padding')
parser.add_argument('--quiet',
action='store_true',
default=False,
help='silences console output')
if len(sys.argv) == 1:
parser.print_usage(sys.stderr)
sys.exit(1)
args = parser.parse_args()
args.images = not args.no_images
args.report = not args.no_report
args.labels = not args.no_labels
# Set up console output
verbose_progress(not args.quiet)
# Speed up pytorch
torch.backends.cudnn.benchmark = True
# Special case: download flag without model to test.
if args.model is None and args.download:
from netdissect.broden import ensure_broden_downloaded
for resolution in [224, 227, 384]:
ensure_broden_downloaded(args.segments, resolution,
args.broden_version)
sys.exit(0)
# Help if broden is not present
if not os.path.isdir(args.segments):
print_progress('Segmentation dataset not found at %s.' % args.segments)
print_progress('Specify dataset directory using --segments [DIR]')
print_progress('To download Broden, run: netdissect --download')
sys.exit(1)
# Construct the network
if args.model is None:
print_progress('No model specified')
sys.exit(1)
# Set up CUDA
args.cuda = not args.no_cuda and torch.cuda.is_available()
if args.cuda:
torch.backends.cudnn.benchmark = True
model = eval_constructor(args.model)
# Default add_scale_offset only for AlexNet-looking models.
if args.add_scale_offset is None:
args.add_scale_offset = ('Alex' in model.__class__.__name__)
# Load its state dict
meta = {}
if args.pthfile is None:
print_progress('Dissecting model without pth file.')
else:
data = torch.load(args.pthfile)
if 'state_dict' in data:
meta = {}
for key in data:
if isinstance(data[key], numbers.Number):
meta[key] = data[key]
data = data['state_dict']
model.load_state_dict(data)
# Update any metadata from files, if any
if args.meta:
for mfilename in args.meta:
with open(mfilename) as f:
meta.update(json.load(f))
# Instrument it and prepare it for eval
if not args.layers:
print_progress('No layers specified')
sys.exit(1)
retain_layers(model, args.layers, args.add_scale_offset)
model.eval()
if args.cuda:
model.cuda()
# Set up the output directory, verify write access
if args.outdir is None:
args.outdir = os.path.join('dissect', type(model).__name__)
print_progress('Writing output into %s.' % args.outdir)
os.makedirs(args.outdir, exist_ok=True)
# Load perturbation
perturbation = numpy.load(args.perturbation) if args.perturbation else None
# Load broden dataset
ds = try_to_load_broden(args.segments, args.imgsize, args.broden_version,
perturbation, args.download, args.size)
if ds is None:
ds = try_to_load_multiseg(args.segments, args.imgsize, perturbation,
args.size)
if ds is None:
print_progress('No segmentation dataset found in %s' % args.segements)
print_progress('use --download to download Broden.')
sys.exit(1)
# Run dissect
dissect(args.outdir,
model,
ds,
recover_image=ReverseNormalize(IMAGE_MEAN, IMAGE_STDEV),
examples_per_unit=args.examples,
netname=args.netname,
meta=meta,
make_images=args.images,
make_labels=args.labels,
make_report=args.report,
make_single_images=args.single_images,
batch_size=args.batch_size,
num_workers=args.num_workers,
settings=vars(args))
