@@ -124,6 +124,7 @@ class DynamiCrafterProcessor:
outs = []
for i in range(batch_size):
model_in_kwargs['transformer_options']['cond_idx'] = i
x_out = apply_model(
x_in[[i]],
t=torch.cat([t[:1]]),
The note is not visible to the blocked user.