a-One-Fan/DL-Art-School
1
0
Fork 0
forked from mrq/DL-Art-School
Code Pull Requests Activity

Assert that temperature is set properly in eval mode.

Browse Source
This commit is contained in:
James Betker 2020-07-22 20:50:59 -06:00
parent c74b9ee2e4
commit 106b8da315
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
codes/models/archs/SwitchedResidualGenerator_arch.py
View File

@ -138,9 +138,9 @@ class ConfigurableSwitchComputer(nn.Module):
x = self.pre_transform(x) x = self.pre_transform(x)
xformed = [t.forward(x) for t in self.transforms] xformed = [t.forward(x) for t in self.transforms]
m = self.multiplexer(identity) m = self.multiplexer(identity)
outputs, attention = self.switch(xformed, m, True) outputs, attention = self.switch(xformed, m, True)
outputs = identity + outputs * self.switch_scale * fixed_scale outputs = identity + outputs * self.switch_scale * fixed_scale
outputs = outputs + self.post_switch_conv(outputs) * self.psc_scale * fixed_scale outputs = outputs + self.post_switch_conv(outputs) * self.psc_scale * fixed_scale
@ -186,6 +186,10 @@ class ConfigurableSwitchedResidualGenerator2(nn.Module):
assert self.upsample_factor == 2 or self.upsample_factor == 4 assert self.upsample_factor == 2 or self.upsample_factor == 4
def forward(self, x): def forward(self, x):
# This is a common bug when evaluating SRG2 generators. It needs to be configured properly in eval mode. Just fail.
if not self.train:
assert self.switches[0].switch.temperature == 1
x = self.initial_conv(x) x = self.initial_conv(x)
self.attentions = [] self.attentions = []