diffusers/pr_12595/en/api/models/autoencoderkl_mochi.md

AutoencoderKLMochi

The 3D variational autoencoder (VAE) model with KL loss used in Mochi was introduced in Mochi 1 Preview by Tsinghua University & ZhipuAI.

The model can be loaded with the following code snippet.

from diffusers import AutoencoderKLMochi

vae = AutoencoderKLMochi.from_pretrained("genmo/mochi-1-preview", subfolder="vae", torch_dtype=torch.float32).to("cuda")

AutoencoderKLMochi[[diffusers.AutoencoderKLMochi]]

class diffusers.AutoencoderKLMochidiffusers.AutoencoderKLMochihttps://github.com/huggingface/diffusers/blob/vr_12595/src/diffusers/models/autoencoders/autoencoder_kl_mochi.py#L660[{"name": "in_channels", "val": ": int = 15"}, {"name": "out_channels", "val": ": int = 3"}, {"name": "encoder_block_out_channels", "val": ": typing.Tuple[int] = (64, 128, 256, 384)"}, {"name": "decoder_block_out_channels", "val": ": typing.Tuple[int] = (128, 256, 512, 768)"}, {"name": "latent_channels", "val": ": int = 12"}, {"name": "layers_per_block", "val": ": typing.Tuple[int, ...] = (3, 3, 4, 6, 3)"}, {"name": "act_fn", "val": ": str = 'silu'"}, {"name": "temporal_expansions", "val": ": typing.Tuple[int, ...] = (1, 2, 3)"}, {"name": "spatial_expansions", "val": ": typing.Tuple[int, ...] = (2, 2, 2)"}, {"name": "add_attention_block", "val": ": typing.Tuple[bool, ...] = (False, True, True, True, True)"}, {"name": "latents_mean", "val": ": typing.Tuple[float, ...] = (-0.06730895953510081, -0.038011381506090416, -0.07477820912866141, -0.05565264470995561, 0.012767231469026969, -0.04703542746246419, 0.043896967884726704, -0.09346305707025976, -0.09918314763016893, -0.008729793427399178, -0.011931556316503654, -0.0321993391887285)"}, {"name": "latents_std", "val": ": typing.Tuple[float, ...] = (0.9263795028493863, 0.9248894543193766, 0.9393059390890617, 0.959253732819592, 0.8244560132752793, 0.917259975397747, 0.9294154431013696, 1.3720942357788521, 0.881393668867029, 0.9168315692124348, 0.9185249279345552, 0.9274757570805041)"}, {"name": "scaling_factor", "val": ": float = 1.0"}]- in_channels (int, optional, defaults to 3) -- Number of channels in the input image.

• out_channels (int, optional, defaults to 3) -- Number of channels in the output.
• block_out_channels (Tuple[int], optional, defaults to (64,)) -- Tuple of block output channels.
• act_fn (str, optional, defaults to "silu") -- The activation function to use.
• scaling_factor (float, optional, defaults to 1.15258426) -- The component-wise standard deviation of the trained latent space computed using the first batch of the training set. This is used to scale the latent space to have unit variance when training the diffusion model. The latents are scaled with the formula z = z * scaling_factor before being passed to the diffusion model. When decoding, the latents are scaled back to the original scale with the formula: z = 1 / scaling_factor * z. For more details, refer to sections 4.3.2 and D.1 of the High-Resolution Image Synthesis with Latent Diffusion Models paper.0

A VAE model with KL loss for encoding images into latents and decoding latent representations into images. Used in Mochi 1 preview.

This model inherits from ModelMixin. Check the superclass documentation for it's generic methods implemented for all models (such as downloading or saving).

wrapperdiffusers.AutoencoderKLMochi.decodehttps://github.com/huggingface/diffusers/blob/vr_12595/src/diffusers/utils/accelerate_utils.py#L43[{"name": "*args", "val": ""}, {"name": "**kwargs", "val": ""}]

enable_tilingdiffusers.AutoencoderKLMochi.enable_tilinghttps://github.com/huggingface/diffusers/blob/vr_12595/src/diffusers/models/autoencoders/autoencoder_kl_mochi.py#L791[{"name": "tile_sample_min_height", "val": ": typing.Optional[int] = None"}, {"name": "tile_sample_min_width", "val": ": typing.Optional[int] = None"}, {"name": "tile_sample_stride_height", "val": ": typing.Optional[float] = None"}, {"name": "tile_sample_stride_width", "val": ": typing.Optional[float] = None"}]- tile_sample_min_height (int, optional) -- The minimum height required for a sample to be separated into tiles across the height dimension.

• tile_sample_min_width (int, optional) -- The minimum width required for a sample to be separated into tiles across the width dimension.
• tile_sample_stride_height (int, optional) -- The minimum amount of overlap between two consecutive vertical tiles. This is to ensure that there are no tiling artifacts produced across the height dimension.
• tile_sample_stride_width (int, optional) -- The stride between two consecutive horizontal tiles. This is to ensure that there are no tiling artifacts produced across the width dimension.0

Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow processing larger images.

tiled_decodediffusers.AutoencoderKLMochi.tiled_decodehttps://github.com/huggingface/diffusers/blob/vr_12595/src/diffusers/models/autoencoders/autoencoder_kl_mochi.py#L1016[{"name": "z", "val": ": Tensor"}, {"name": "return_dict", "val": ": bool = True"}]- z (torch.Tensor) -- Input batch of latent vectors.

• return_dict (bool, optional, defaults to True) -- Whether or not to return a ~models.vae.DecoderOutput instead of a plain tuple.0~models.vae.DecoderOutput or tupleIf return_dict is True, a ~models.vae.DecoderOutput is returned, otherwise a plain tuple is returned.

Decode a batch of images using a tiled decoder.

tiled_encodediffusers.AutoencoderKLMochi.tiled_encodehttps://github.com/huggingface/diffusers/blob/vr_12595/src/diffusers/models/autoencoders/autoencoder_kl_mochi.py#L959[{"name": "x", "val": ": Tensor"}]- x (torch.Tensor) -- Input batch of videos.0torch.TensorThe latent representation of the encoded videos. Encode a batch of images using a tiled encoder.

DecoderOutput[[diffusers.models.autoencoders.vae.DecoderOutput]]

class diffusers.models.autoencoders.vae.DecoderOutputdiffusers.models.autoencoders.vae.DecoderOutputhttps://github.com/huggingface/diffusers/blob/vr_12595/src/diffusers/models/autoencoders/vae.py#L47[{"name": "sample", "val": ": Tensor"}, {"name": "commit_loss", "val": ": typing.Optional[torch.FloatTensor] = None"}]- sample (torch.Tensor of shape (batch_size, num_channels, height, width)) -- The decoded output sample from the last layer of the model.0

Output of decoding method.