steffen/masks_augments_compare-21
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Debugging und hparam als Mapping

Browse Source
This commit is contained in:
Si11ium
2020-04-24 17:55:43 +02:00
parent f607324606
commit a3c9f7eac8
1 changed files with 23 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
models/binary_classifier.py
View File

@@ -3,7 +3,7 @@ from torch import nn
from torch.optim import Adam from torch.optim import Adam
from ml_lib.modules.blocks import ConvModule from ml_lib.modules.blocks import ConvModule
from ml_lib.modules.utils import LightningBaseModule from ml_lib.modules.utils import LightningBaseModule, Flatten
class BinaryClassifier(LightningBaseModule): class BinaryClassifier(LightningBaseModule):
@@ -12,7 +12,7 @@ class BinaryClassifier(LightningBaseModule):
return cls.__name__ return cls.__name__
def configure_optimizers(self): def configure_optimizers(self):
return Adam(lr=self.hparams.train.lr) return Adam(params=self.Parameters, lr=self.hparams.train.lr)
def training_step(self, batch_xy, batch_nb, *args, **kwargs): def training_step(self, batch_xy, batch_nb, *args, **kwargs):
batch_x, batch_y = batch_xy batch_x, batch_y = batch_xy
@@ -38,14 +38,29 @@ class BinaryClassifier(LightningBaseModule):
self.criterion = nn.BCELoss() self.criterion = nn.BCELoss()
# Additional parameters # Additional parameters
self.in_shape = self.hparams.model_params.in_shape self.in_shape = self.hparams.in_shape
# Model Modules # Model Modules
self.conv_1 = ConvModule(self.in_shape, 32, 5, conv_stride=4, **hparams.model_params) self.conv_1 = ConvModule(self.in_shape, 32, 5, conv_stride=4, **hparams)
self.conv_2 = ConvModule(self.conv_1.shape, 64, 7, conv_stride=2, **hparams.model_params) self.conv_2 = ConvModule(self.conv_1.shape, 64, 7, conv_stride=2, **hparams)
self.conv_3 = ConvModule(self.conv_1.shape, 128, 9, conv_stride=2, **hparams.model_params) self.conv_3 = ConvModule(self.conv_2.shape, 128, 9, conv_stride=2, **hparams)
self.flat = Flatten(self.conv_3.shape)
self.full_1 = nn.Linear(self.flat.shape, 32)
self.full_2 = nn.Linear(self.full_1.out_features, self.full_1.out_features // 2)
self.activation = self.hparams.activation()
self.full_out = nn.Linear(self.full_2.out_features, 2)
self.sigmoid = nn.Sigmoid()
def forward(self, batch, **kwargs): def forward(self, batch, **kwargs):
tensor = self.conv_1(batch)
tensor = self.conv_2(tensor)
tensor = self.conv_3(tensor)
tensor = self.full_1(tensor)
tensor = self.activation(tensor)
tensor = self.full_2(tensor)
tensor = self.activation(tensor)
tensor = self.full_out(tensor)
tensor = self.sigmoid(tensor)
return batch return batch