Hugging Face's logo

Diffusers documentation

JoyImageEditTransformer3DModel

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v0.42.1).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

JoyImageEditTransformer3DModel

Model dapat dimuat dengan potongan kode berikut.

from diffusers import JoyImageEditTransformer3DModel

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

JoyImageEditTransformer3DModel

kelas diffusers.JoyImageEditTransformer3DModel

< >

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

Model JoyImage Transformer untuk pembuatan / pengeditan gambar.

Arsitektur DiT dual-stream dengan embedding kondisioning bergaya WAN dan embedding posisi rotasi kustom.

forward

< >

( hidden_states: Tensortimestep: Tensorencoder_hidden_states: Tensor = Nonereturn_dict: bool = True )

Parameter

  • hidden_states (torch.Tensor dengan bentuk (batch_size, num_channels, num_frames, height, width) atau (batch_size, num_items, num_channels, num_frames, height, width)) — hidden_states masukan.
  • timestep (torch.LongTensor) — Digunakan untuk menunjukkan langkah denoising.
  • encoder_hidden_states (torch.Tensor, opsional) — Embedding kondisional (embedding yang dihitung dari kondisi input seperti prompt) yang akan digunakan.
  • return_dict (bool, opsional, default True) — Apakah akan mengembalikan ~models.transformer_2d.Transformer2DModelOutput alih-alih tuple biasa.

Metode forward dari JoyImageEditTransformer3DModel.

Transformer2DModelOutput

kelas diffusers.models.modeling_outputs.Transformer2DModelOutput

< >

( sample: torch.Tensor )

Parameter

  • sample (torch.Tensor dengan bentuk (batch_size, num_channels, height, width) atau (batch size, num_vector_embeds - 1, num_latent_pixels) jika Transformer2DModel bersifat diskrit) — Output hidden states yang dikondisikan pada input encoder_hidden_states. Jika diskrit, mengembalikan distribusi probabilitas untuk latent pixels yang tidak berisik.

Output dari Transformer2DModel.

Update on GitHub

←JoyImageEditPlusTransformer3DModel Kandinsky 6 Transformers→