re-derive sqrt alpha bar and sqrt one minus alphabar

This is the only place these values are ever referenced outside of training code so this change is very justifiable and more consistent.

1 files changed, 1 insertions, 1 deletions

diff --git a/modules/sd_samplers_timesteps.py b/modules/sd_samplers_timesteps.py
index b17a8f93..c4bd5c12 100644
--- a/modules/sd_samplers_timesteps.py
+++ b/modules/sd_samplers_timesteps.py
@@ -36,7 +36,7 @@ class CompVisTimestepsVDenoiser(torch.nn.Module):
         self.inner_model = model

 

     def predict_eps_from_z_and_v(self, x_t, t, v):

-        return self.inner_model.sqrt_alphas_cumprod[t.to(torch.int), None, None, None] * v + self.inner_model.sqrt_one_minus_alphas_cumprod[t.to(torch.int), None, None, None] * x_t

+        return torch.sqrt(self.inner_model.alphas_cumprod)[t.to(torch.int), None, None, None] * v + torch.sqrt(1 - self.inner_model.alphas_cumprod)[t.to(torch.int), None, None, None] * x_t

 

     def forward(self, input, timesteps, **kwargs):

         model_output = self.inner_model.apply_model(input, timesteps, **kwargs)