From a09b6499f8035e70c7b4f7afe995f3dba9db4e9a Mon Sep 17 00:00:00 2001 From: Alex Brooks Date: Mon, 2 Mar 2026 09:50:08 -0700 Subject: [PATCH] [Bugfix][Model] Fix LongCat Image Config Handling / Layer Creation (#1485) Signed-off-by: Alex Brooks Co-authored-by: SYLAR <125541396+lishunyang12@users.noreply.github.com> --- .../longcat_image_transformer.py | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) diff --git a/vllm_omni/diffusion/models/longcat_image/longcat_image_transformer.py b/vllm_omni/diffusion/models/longcat_image/longcat_image_transformer.py index 2f6d54aef..329fc648b 100644 --- a/vllm_omni/diffusion/models/longcat_image/longcat_image_transformer.py +++ b/vllm_omni/diffusion/models/longcat_image/longcat_image_transformer.py @@ -517,17 +517,18 @@ class LongCatImageTransformer2DModel(nn.Module): def __init__( self, od_config: OmniDiffusionConfig, - patch_size: int = 1, - in_channels: int = 64, - num_layers: int = 19, - num_single_layers: int = 38, - attention_head_dim: int = 128, - num_attention_heads: int = 24, - joint_attention_dim: int = 3584, - pooled_projection_dim: int = 3584, - axes_dims_rope: list[int] = [16, 56, 56], ): super().__init__() + model_config = od_config.tf_model_config + patch_size = model_config.patch_size + in_channels = model_config.in_channels + num_layers = model_config.num_layers + num_single_layers = model_config.num_single_layers + attention_head_dim = model_config.attention_head_dim + num_attention_heads = model_config.num_attention_heads + joint_attention_dim = model_config.joint_attention_dim + pooled_projection_dim = model_config.pooled_projection_dim + axes_dims_rope = getattr(model_config, "axes_dims_rope", [16, 56, 56]) self.out_channels = in_channels self.inner_dim = num_attention_heads * attention_head_dim self.pooled_projection_dim = pooled_projection_dim