Hugging Face's logo

Diffusers documentation

QwenImageTransformer2DModel

Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

QwenImageTransformer2DModel

The model can be loaded with the following code snippet.

from diffusers import QwenImageTransformer2DModel

transformer = QwenImageTransformer2DModel.from_pretrained("Qwen/QwenImage-20B", subfolder="transformer", torch_dtype=torch.bfloat16)

QwenImageTransformer2DModel

class diffusers.QwenImageTransformer2DModel

< >

( patch_size: int = 2 in_channels: int = 64 out_channels: typing.Optional[int] = 16 num_layers: int = 60 attention_head_dim: int = 128 num_attention_heads: int = 24 joint_attention_dim: int = 3584 guidance_embeds: bool = False axes_dims_rope: typing.Tuple[int, int, int] = (16, 56, 56) )

Parameters

  • patch_size (int, defaults to 2) — Patch size to turn the input data into small patches.
  • in_channels (int, defaults to 64) — The number of channels in the input.
  • out_channels (int, optional, defaults to None) — The number of channels in the output. If not specified, it defaults to in_channels.
  • num_layers (int, defaults to 60) — The number of layers of dual stream DiT blocks to use.
  • attention_head_dim (int, defaults to 128) — The number of dimensions to use for each attention head.
  • num_attention_heads (int, defaults to 24) — The number of attention heads to use.
  • joint_attention_dim (int, defaults to 3584) — The number of dimensions to use for the joint attention (embedding/channel dimension of encoder_hidden_states).
  • guidance_embeds (bool, defaults to False) — Whether to use guidance embeddings for guidance-distilled variant of the model.
  • axes_dims_rope (Tuple[int], defaults to (16, 56, 56)) — The dimensions to use for the rotary positional embeddings.

The Transformer model introduced in Qwen.

forward

< >

( hidden_states: Tensor encoder_hidden_states: Tensor = None encoder_hidden_states_mask: Tensor = None timestep: LongTensor = None img_shapes: typing.Optional[typing.List[typing.Tuple[int, int, int]]] = None txt_seq_lens: typing.Optional[typing.List[int]] = None guidance: Tensor = None attention_kwargs: typing.Optional[typing.Dict[str, typing.Any]] = None return_dict: bool = True )

Parameters

  • hidden_states (torch.Tensor of shape (batch_size, image_sequence_length, in_channels)) — Input hidden_states.
  • encoder_hidden_states (torch.Tensor of shape (batch_size, text_sequence_length, joint_attention_dim)) — Conditional embeddings (embeddings computed from the input conditions such as prompts) to use.
  • encoder_hidden_states_mask (torch.Tensor of shape (batch_size, text_sequence_length)) — Mask of the input conditions.
  • timestep ( torch.LongTensor) — Used to indicate denoising step.
  • attention_kwargs (dict, optional) — A kwargs dictionary that if specified is passed along to the AttentionProcessor as defined under self.processor in diffusers.models.attention_processor.
  • return_dict (bool, optional, defaults to True) — Whether or not to return a ~models.transformer_2d.Transformer2DModelOutput instead of a plain tuple.

The QwenTransformer2DModel forward method.

Transformer2DModelOutput

class diffusers.models.modeling_outputs.Transformer2DModelOutput

< >

( sample: torch.Tensor )

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.

< > Update on GitHub

←PriorTransformer SanaTransformer2DModel→