In [1]:
%reload_ext autoreload
%autoreload 2
%matplotlib inline
In [7]:
from fastai.conv_learner import *
from fastai.models import darknet
from pathlib import Path
In [8]:
PATH = Path('data/imagenet')
PATH_TRAIN = PATH/'train'
In [5]:
# sz = 256
# bs = 32
# darknet53 = darknet.darknet_53()
# tfms = tfms_from_model(darknet53, sz)
# model_data = ImageClassifierData.from_paths(PATH, bs=bs, tfms=tfms, val_name='train')
# learner = ConvLearner.from_model_data(darknet53, model_data)
In [9]:
learner.crit
Out[9]:
<function torch.nn.functional.nll_loss(input, target, weight=None, size_average=True, ignore_index=-100, reduce=True)>
In [12]:
model_data
Out[12]:
<fastai.dataset.ImageClassifierData at 0x7ff722707518>
In [13]:
learner.crit
Out[13]:
<function torch.nn.functional.nll_loss(input, target, weight=None, size_average=True, ignore_index=-100, reduce=True)>
In [14]:
learner.crit = F.cross_entropy
In [15]:
learner.crit
Out[15]:
<function torch.nn.functional.cross_entropy(input, target, weight=None, size_average=True, ignore_index=-100, reduce=True)>
Checking behavior of .pretrained in fastai/conv_learner.py. Looking at number of FC layers at the end of a stock PyTorch model, compared to loading through FastAI:
In [9]:
from torchvision.models import resnet18
In [10]:
resnet18
Out[10]:
<function torchvision.models.resnet.resnet18(pretrained=False, **kwargs)>
In [11]:
resnet18 = resnet18()
In [3]:
# resnet18
1 linear layer at the end.
In [18]:
from fastai.conv_learner import *
In [19]:
resnet18
Out[19]:
<function torchvision.models.resnet.resnet18(pretrained=False, **kwargs)>
What fastai originally imports is the resnet18 constructor function from torchvision.models.resnet. Good to know.
In [23]:
learner = ConvLearner.from_model_data(resnet18(), model_data)
In [2]:
# learner
In [20]:
from pathlib import Path
PATH = Path('data/cifar10')
In [21]:
sz = 32
bs = 64
tfms = tfms_from_model(resnet18, sz)
model_data = ImageClassifierData.from_paths(PATH, bs, tfms=tfms, val_name='test')
In [10]:
learner = ConvLearner.pretrained(resnet18, model_data)
Downloading: "https://download.pytorch.org/models/resnet18-5c106cde.pth" to /home/ubuntu/.torch/models/resnet18-5c106cde.pth
100%|██████████| 46827520/46827520 [00:00<00:00, 65104058.86it/s]
In [1]:
# learner
Right. fastai strips the FC layer and its associated pooling layers, and replaces them with adaptive pooling and 2 linear layers with batchNorm and dropout; with a LogSoftmax output layer.
In [12]:
from fastai.models import darknet
In [13]:
darknet.darknet_53
Out[13]:
<function fastai.models.darknet.darknet_53(num_classes=1000)>
In [14]:
tfms = tfms_from_stats(imagenet_stats, sz)
model_data = ImageClassifierData.from_paths(PATH, bs=bs, tfms=tfms, val_name='test')
darknet53 = darknet.darknet_53
In [4]:
# learner = ConvLearner.from_model_data(darknet53(num_classes=10), model_data)
# learner
In [5]:
# learner = ConvLearner.pretrained(darknet53(num_classes=10), model_data)
# learner
In [4]:
%matplotlib inline
%reload_ext autoreload
%autoreload 2
from fastai.conv_learner import *
from fastai.models import darknet
from pathlib import Path
PATH = Path('data/imagenet')
PATH_TRAIN = PATH/'train'
In [5]:
bs = 32
sz = 256
tfms = tfms_from_stats(imagenet_stats, sz)
model_data = ImageClassifierData.from_paths(PATH, bs=bs, tfms=tfms)
darknet53 = darknet.darknet_53
.from_model_data
In [7]:
learner = ConvLearner.from_model_data(darknet53(), model_data)
In [11]:
# learner.summary()
head:
('AdaptiveAvgPool2d-232',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 1, 1]),
('nb_params', 0)])),
('Flatten-233',
OrderedDict([('input_shape', [-1, 1024, 1, 1]),
('output_shape', [-1, 1024]),
('nb_params', 0)])),
('Linear-234',
OrderedDict([('input_shape', [-1, 1024]),
('output_shape', [-1, 1000]),
('trainable', True),
('nb_params', 1025000)]))])learner with NLL loss fails Learning-Rate Finder phase due to negative losses:
In [14]:
learner.lr_find()
learner.sched.plot()
4%|▎ | 18/487 [01:26<37:32, 4.80s/it, loss=-0.0191]
In [15]:
learner.crit
Out[15]:
<function torch.nn.functional.nll_loss(input, target, weight=None, size_average=True, ignore_index=-100, reduce=True)>
.pretrained
In [17]:
learner = ConvLearner.pretrained(darknet53(), model_data)
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-17-66ef9014e32e> in <module>()
----> 1 learner = ConvLearner.pretrained(darknet53(), model_data)
~/Kaukasos/FADL2/fastai/conv_learner.py in pretrained(cls, f, data, ps, xtra_fc, xtra_cut, custom_head, precompute, pretrained, **kwargs)
110 pretrained=True, **kwargs):
111 models = ConvnetBuilder(f, data.c, data.is_multi, data.is_reg,
--> 112 ps=ps, xtra_fc=xtra_fc, xtra_cut=xtra_cut, custom_head=custom_head, pretrained=pretrained)
113 return cls(data, models, precompute, **kwargs)
114
~/Kaukasos/FADL2/fastai/conv_learner.py in __init__(self, f, c, is_multi, is_reg, ps, xtra_fc, xtra_cut, custom_head, pretrained)
38 else: cut,self.lr_cut = 0,0
39 cut-=xtra_cut
---> 40 layers = cut_model(f(pretrained), cut)
41 self.nf = model_features[f] if f in model_features else (num_features(layers)*2)
42 if not custom_head: layers += [AdaptiveConcatPool2d(), Flatten()]
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
355 result = self._slow_forward(*input, **kwargs)
356 else:
--> 357 result = self.forward(*input, **kwargs)
358 for hook in self._forward_hooks.values():
359 hook_result = hook(self, input, result)
~/Kaukasos/FADL2/fastai/models/darknet.py in forward(self, x)
42 self.layers = nn.Sequential(*layers)
43
---> 44 def forward(self, x): return self.layers(x)
45
46 def darknet_53(num_classes=1000): return Darknet([1,2,8,8,4], num_classes)
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
355 result = self._slow_forward(*input, **kwargs)
356 else:
--> 357 result = self.forward(*input, **kwargs)
358 for hook in self._forward_hooks.values():
359 hook_result = hook(self, input, result)
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/modules/container.py in forward(self, input)
65 def forward(self, input):
66 for module in self._modules.values():
---> 67 input = module(input)
68 return input
69
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
355 result = self._slow_forward(*input, **kwargs)
356 else:
--> 357 result = self.forward(*input, **kwargs)
358 for hook in self._forward_hooks.values():
359 hook_result = hook(self, input, result)
~/Kaukasos/FADL2/fastai/models/darknet.py in forward(self, x)
13 self.relu = nn.LeakyReLU(0.1, inplace=True)
14
---> 15 def forward(self, x): return self.relu(self.bn(self.conv(x)))
16
17 class DarknetBlock(nn.Module):
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
355 result = self._slow_forward(*input, **kwargs)
356 else:
--> 357 result = self.forward(*input, **kwargs)
358 for hook in self._forward_hooks.values():
359 hook_result = hook(self, input, result)
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/modules/conv.py in forward(self, input)
280 def forward(self, input):
281 return F.conv2d(input, self.weight, self.bias, self.stride,
--> 282 self.padding, self.dilation, self.groups)
283
284
~/src/anaconda3/envs/fastai/lib/python3.6/site-packages/torch/nn/functional.py in conv2d(input, weight, bias, stride, padding, dilation, groups)
82 >>> F.conv2d(inputs, filters, padding=1)
83 """
---> 84 if input is not None and input.dim() != 4:
85 raise ValueError("Expected 4D tensor as input, got {}D tensor instead.".format(input.dim()))
86
AttributeError: 'bool' object has no attribute 'dim'
In [ ]:
learner.summary()
In [ ]:
learner.lr_find()
learner.sched.plot()
In [1]:
from pathlib import Path
from fastai.conv_learner import *
from fastai.models import darknet
In [2]:
PATH = Path('data/imagenet')
sz = 256; bs=32
tfms = tfms_from_stats(imagenet_stats, sz)
model_data = ImageClassifierData.from_paths(PATH, bs=bs, tfms=tfms, val_name='train')
darknet53 = darknet.darknet_53
In [3]:
learner = ConvLearner.from_model_data(darknet53(), model_data)
In [4]:
# learner.summary()
In [5]:
learner.lr_find()
learner.sched.plot()
92%|█████████▏| 448/487 [33:30<02:55, 4.49s/it, loss=28.3]
In [6]:
learner.summary()
Out[6]:
OrderedDict([('Conv2d-1',
OrderedDict([('input_shape', [-1, 3, 256, 256]),
('output_shape', [-1, 32, 256, 256]),
('trainable', True),
('nb_params', 864)])),
('BatchNorm2d-2',
OrderedDict([('input_shape', [-1, 32, 256, 256]),
('output_shape', [-1, 32, 256, 256]),
('trainable', True),
('nb_params', 64)])),
('LeakyReLU-3',
OrderedDict([('input_shape', [-1, 32, 256, 256]),
('output_shape', [-1, 32, 256, 256]),
('nb_params', 0)])),
('ConvBN-4',
OrderedDict([('input_shape', [-1, 3, 256, 256]),
('output_shape', [-1, 32, 256, 256]),
('nb_params', 0)])),
('Conv2d-5',
OrderedDict([('input_shape', [-1, 32, 256, 256]),
('output_shape', [-1, 64, 127, 127]),
('trainable', True),
('nb_params', 18432)])),
('BatchNorm2d-6',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('trainable', True),
('nb_params', 128)])),
('LeakyReLU-7',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('nb_params', 0)])),
('ConvBN-8',
OrderedDict([('input_shape', [-1, 32, 256, 256]),
('output_shape', [-1, 64, 127, 127]),
('nb_params', 0)])),
('Conv2d-9',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 32, 127, 127]),
('trainable', True),
('nb_params', 2048)])),
('BatchNorm2d-10',
OrderedDict([('input_shape', [-1, 32, 127, 127]),
('output_shape', [-1, 32, 127, 127]),
('trainable', True),
('nb_params', 64)])),
('LeakyReLU-11',
OrderedDict([('input_shape', [-1, 32, 127, 127]),
('output_shape', [-1, 32, 127, 127]),
('nb_params', 0)])),
('ConvBN-12',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 32, 127, 127]),
('nb_params', 0)])),
('Conv2d-13',
OrderedDict([('input_shape', [-1, 32, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('trainable', True),
('nb_params', 18432)])),
('BatchNorm2d-14',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('trainable', True),
('nb_params', 128)])),
('LeakyReLU-15',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('nb_params', 0)])),
('ConvBN-16',
OrderedDict([('input_shape', [-1, 32, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('nb_params', 0)])),
('DarknetBlock-17',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 64, 127, 127]),
('nb_params', 0)])),
('Conv2d-18',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 128, 125, 125]),
('trainable', True),
('nb_params', 73728)])),
('BatchNorm2d-19',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-20',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('ConvBN-21',
OrderedDict([('input_shape', [-1, 64, 127, 127]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('Conv2d-22',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('trainable', True),
('nb_params', 8192)])),
('BatchNorm2d-23',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('trainable', True),
('nb_params', 128)])),
('LeakyReLU-24',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('nb_params', 0)])),
('ConvBN-25',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('nb_params', 0)])),
('Conv2d-26',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('trainable', True),
('nb_params', 73728)])),
('BatchNorm2d-27',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-28',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('ConvBN-29',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('DarknetBlock-30',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('Conv2d-31',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('trainable', True),
('nb_params', 8192)])),
('BatchNorm2d-32',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('trainable', True),
('nb_params', 128)])),
('LeakyReLU-33',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('nb_params', 0)])),
('ConvBN-34',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 64, 125, 125]),
('nb_params', 0)])),
('Conv2d-35',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('trainable', True),
('nb_params', 73728)])),
('BatchNorm2d-36',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-37',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('ConvBN-38',
OrderedDict([('input_shape', [-1, 64, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('DarknetBlock-39',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 128, 125, 125]),
('nb_params', 0)])),
('Conv2d-40',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-41',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-42',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-43',
OrderedDict([('input_shape', [-1, 128, 125, 125]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-44',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-45',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-46',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-47',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-48',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-49',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-50',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-51',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-52',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-53',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-54',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-55',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-56',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-57',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-58',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-59',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-60',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-61',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-62',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-63',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-64',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-65',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-66',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-67',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-68',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-69',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-70',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-71',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-72',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-73',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-74',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-75',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-76',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-77',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-78',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-79',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-80',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-81',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-82',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-83',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-84',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-85',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-86',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-87',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-88',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-89',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-90',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-91',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-92',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-93',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-94',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-95',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-96',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-97',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-98',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-99',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-100',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-101',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-102',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-103',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-104',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-105',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-106',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-107',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 32768)])),
('BatchNorm2d-108',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('trainable', True),
('nb_params', 256)])),
('LeakyReLU-109',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('ConvBN-110',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 128, 62, 62]),
('nb_params', 0)])),
('Conv2d-111',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 294912)])),
('BatchNorm2d-112',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-113',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('ConvBN-114',
OrderedDict([('input_shape', [-1, 128, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('DarknetBlock-115',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 256, 62, 62]),
('nb_params', 0)])),
('Conv2d-116',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-117',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-118',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-119',
OrderedDict([('input_shape', [-1, 256, 62, 62]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-120',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-121',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-122',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-123',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-124',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-125',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-126',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-127',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-128',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-129',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-130',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-131',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-132',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-133',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-134',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-135',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-136',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-137',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-138',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-139',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-140',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-141',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-142',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-143',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-144',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-145',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-146',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-147',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-148',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-149',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-150',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-151',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-152',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-153',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-154',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-155',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-156',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-157',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-158',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-159',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-160',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-161',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-162',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-163',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-164',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-165',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-166',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-167',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-168',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-169',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-170',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-171',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-172',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-173',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-174',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-175',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-176',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-177',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-178',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-179',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-180',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-181',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-182',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-183',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 131072)])),
('BatchNorm2d-184',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('trainable', True),
('nb_params', 512)])),
('LeakyReLU-185',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('ConvBN-186',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 256, 30, 30]),
('nb_params', 0)])),
('Conv2d-187',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1179648)])),
('BatchNorm2d-188',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-189',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('ConvBN-190',
OrderedDict([('input_shape', [-1, 256, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('DarknetBlock-191',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 512, 30, 30]),
('nb_params', 0)])),
('Conv2d-192',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 4718592)])),
('BatchNorm2d-193',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 2048)])),
('LeakyReLU-194',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('ConvBN-195',
OrderedDict([('input_shape', [-1, 512, 30, 30]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('Conv2d-196',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 524288)])),
('BatchNorm2d-197',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-198',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('ConvBN-199',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('Conv2d-200',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 4718592)])),
('BatchNorm2d-201',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 2048)])),
('LeakyReLU-202',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('ConvBN-203',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('DarknetBlock-204',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('Conv2d-205',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 524288)])),
('BatchNorm2d-206',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-207',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('ConvBN-208',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('Conv2d-209',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 4718592)])),
('BatchNorm2d-210',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 2048)])),
('LeakyReLU-211',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('ConvBN-212',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('DarknetBlock-213',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('Conv2d-214',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 524288)])),
('BatchNorm2d-215',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-216',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('ConvBN-217',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('Conv2d-218',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 4718592)])),
('BatchNorm2d-219',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 2048)])),
('LeakyReLU-220',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('ConvBN-221',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('DarknetBlock-222',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('Conv2d-223',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 524288)])),
('BatchNorm2d-224',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('trainable', True),
('nb_params', 1024)])),
('LeakyReLU-225',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('ConvBN-226',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 512, 14, 14]),
('nb_params', 0)])),
('Conv2d-227',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 4718592)])),
('BatchNorm2d-228',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('trainable', True),
('nb_params', 2048)])),
('LeakyReLU-229',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('ConvBN-230',
OrderedDict([('input_shape', [-1, 512, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('DarknetBlock-231',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 14, 14]),
('nb_params', 0)])),
('AdaptiveAvgPool2d-232',
OrderedDict([('input_shape', [-1, 1024, 14, 14]),
('output_shape', [-1, 1024, 1, 1]),
('nb_params', 0)])),
('Flatten-233',
OrderedDict([('input_shape', [-1, 1024, 1, 1]),
('output_shape', [-1, 1024]),
('nb_params', 0)])),
('Linear-234',
OrderedDict([('input_shape', [-1, 1024]),
('output_shape', [-1, 1000]),
('trainable', True),
('nb_params', 1025000)])),
('LogSoftmax-235',
OrderedDict([('input_shape', [-1, 1000]),
('output_shape', [-1, 1000]),
('nb_params', 0)]))])
In [7]:
learner.crit
Out[7]:
<function torch.nn.functional.nll_loss(input, target, weight=None, size_average=True, ignore_index=-100, reduce=True)>
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
Content source: WNoxchi/Kaukasos
Similar notebooks: