diff --git a/src/diffusers/models/attention.py b/src/diffusers/models/attention.py
index 25e1ea28dcf0..f963310f12eb 100644
--- a/src/diffusers/models/attention.py
+++ b/src/diffusers/models/attention.py
@@ -144,10 +144,11 @@ def forward(self, hidden_states, context=None):
         residual = hidden_states
         hidden_states = self.norm(hidden_states)
         hidden_states = self.proj_in(hidden_states)
-        hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(batch, height * weight, channel)
+        inner_dim = hidden_states.shape[1]
+        hidden_states = hidden_states.permute(0, 2, 3, 1).reshape(batch, height * weight, inner_dim)
         for block in self.transformer_blocks:
             hidden_states = block(hidden_states, context=context)
-        hidden_states = hidden_states.reshape(batch, height, weight, channel).permute(0, 3, 1, 2)
+        hidden_states = hidden_states.reshape(batch, height, weight, inner_dim).permute(0, 3, 1, 2)
         hidden_states = self.proj_out(hidden_states)
         return hidden_states + residual