Diffusers documentation
LongCatImageTransformer2DModel
LongCatImageTransformer2DModel
The model can be loaded with the following code snippet.
from diffusers import LongCatImageTransformer2DModel
transformer = LongCatImageTransformer2DModel.from_pretrained("meituan-longcat/LongCat-Image ", subfolder="transformer", dtype=torch.bfloat16)LongCatImageTransformer2DModel
class diffusers.LongCatImageTransformer2DModel
< source >( patch_size: int = 1in_channels: int = 64num_layers: int = 19num_single_layers: int = 38attention_head_dim: int = 128num_attention_heads: int = 24joint_attention_dim: int = 3584pooled_projection_dim: int = 3584axes_dims_rope: list = [16, 56, 56] )
The Transformer model introduced in Longcat-Image.
forward
< source >( hidden_states: Tensorencoder_hidden_states: Tensor = Nonetimestep: LongTensor = Noneimg_ids: Tensor = Nonetxt_ids: Tensor = Noneguidance: Tensor = Nonereturn_dict: bool = True )
Parameters
- hidden_states (
torch.FloatTensorof shape(batch size, channel, height, width)) — Inputhidden_states. - encoder_hidden_states (
torch.FloatTensorof shape(batch size, sequence_len, embed_dims)) — Conditional embeddings (embeddings computed from the input conditions such as prompts) to use. - timestep (
torch.LongTensor) — Used to indicate denoising step. - img_ids (
torch.Tensor) — Image position ids used to compute the rotary positional embeddings. - txt_ids (
torch.Tensor) — Text position ids used to compute the rotary positional embeddings. - guidance (
torch.Tensor, optional) — Guidance scale embedding used for guidance-distilled variants of the model. - return_dict (
bool, optional, defaults toTrue) — Whether or not to return a~models.transformer_2d.Transformer2DModelOutputinstead of a plain tuple.
The forward method.