diffusers/pr_14385/en/api/models/transformer_joyimage.md

JoyImageEditTransformer3DModel

The model can be loaded with the following code snippet.

from diffusers import JoyImageEditTransformer3DModel

transformer = JoyImageEditTransformer3DModel.from_pretrained("jdopensource/JoyAI-Image-Edit-Diffusers", subfolder="transformer", dtype=torch.bfloat16)

JoyImageEditTransformer3DModel[[diffusers.JoyImageEditTransformer3DModel]]

diffusers.JoyImageEditTransformer3DModel[[diffusers.JoyImageEditTransformer3DModel]]

diffusers.JoyImageEditTransformer3DModel(patch_size: list = [1, 2, 2], in_channels: int = 16, out_channels: int | None = None, hidden_size: int = 3072, num_attention_heads: int = 24, text_dim: int = 4096, mlp_width_ratio: float = 4.0, num_layers: int = 20, rope_dim_list: list = [16, 56, 56], rope_type: str = 'rope', theta: int = 256)

Source

JoyImage Transformer model for image generation / editing.

Dual-stream DiT architecture with WAN-style conditioning embeddings and custom rotary position embeddings.

forward[[diffusers.JoyImageEditTransformer3DModel.forward]]

forward(hidden_states: Tensor, timestep: Tensor, encoder_hidden_states: Tensor = None, return_dict: bool = True)

Source

Parameters:

hidden_states (torch.Tensor of shape (batch_size, num_channels, num_frames, height, width) or (batch_size, num_items, num_channels, num_frames, height, width)) : Input hidden_states.

timestep (torch.LongTensor) : Used to indicate denoising step.

encoder_hidden_states (torch.Tensor, optional) : Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.

return_dict (bool, optional, defaults to True) : Whether or not to return a ~models.transformer_2d.Transformer2DModelOutput instead of a plain tuple.

The JoyImageEditTransformer3DModel forward method.

Transformer2DModelOutput[[diffusers.models.modeling_outputs.Transformer2DModelOutput]]

diffusers.models.modeling_outputs.Transformer2DModelOutput[[diffusers.models.modeling_outputs.Transformer2DModelOutput]]

diffusers.models.modeling_outputs.Transformer2DModelOutput(sample: torch.Tensor)

Source

Parameters:

sample (torch.Tensor of shape (batch_size, num_channels, height, width) or (batch size, num_vector_embeds - 1, num_latent_pixels) if Transformer2DModel is discrete) : The hidden states output conditioned on the encoder_hidden_states input. If discrete, returns probability distributions for the unnoised latent pixels.

The output of Transformer2DModel.