Make lumina model work with any latent resolution.

comfyanonymous · Feb 10, 2025 · 4027466 · 4027466
1 parent 095d867
commit 4027466
Showing 1 changed file with 4 additions and 1 deletion.
diff --git a/comfy/ldm/lumina/model.py b/comfy/ldm/lumina/model.py
@@ -6,6 +6,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
+import comfy.ldm.common_dit
 
 from comfy.ldm.modules.diffusionmodules.mmdit import TimestepEmbedder, RMSNorm
 from comfy.ldm.modules.attention import optimized_attention_masked
@@ -594,6 +595,8 @@ def forward(self, x, timesteps, context, num_tokens, attention_mask=None, **kwar
         t = 1.0 - timesteps
         cap_feats = context
         cap_mask = attention_mask
+        bs, c, h, w = x.shape
+        x = comfy.ldm.common_dit.pad_to_patch_size(x, (self.patch_size, self.patch_size))
         """
         Forward pass of NextDiT.
         t: (N,) tensor of diffusion timesteps
@@ -613,7 +616,7 @@ def forward(self, x, timesteps, context, num_tokens, attention_mask=None, **kwar
             x = layer(x, mask, freqs_cis, adaln_input)
 
         x = self.final_layer(x, adaln_input)
-        x = self.unpatchify(x, img_size, cap_size, return_tensor=x_is_tensor)
+        x = self.unpatchify(x, img_size, cap_size, return_tensor=x_is_tensor)[:,:,:h,:w]
 
         return -x