diffusers/pr_13921/en/api/models/cosmos3_omni_transformer.md

Cosmos3OmniTransformer

A Mixture-of-Transformer (MoT) joint vision-language transformer introduced as part of NVIDIA's Cosmos3 world foundation model family. The model runs two parallel computation pathways over a packed joint sequence:

• a causal "understanding" pathway that self-attends over text tokens with causal masking, and
• a bi-directional "generation" pathway that cross-attends from generation tokens (vision + optional sound latents) over the full understanding-plus-generation key/value set.

The two pathways share the same hidden size and number of layers but maintain separate Q/K/V/O projections, MLPs, and RMSNorm parameters, which is what makes the architecture a Mixture-of-Transformer rather than a standard Mixture-of-Experts. Position information is supplied through a 3D multimodal RoPE (mRoPE) that interleaves temporal / height / width frequencies for video latents and reuses the temporal axis for text and audio.

The model can be loaded as follows.

import torch
from diffusers import Cosmos3OmniTransformer

transformer = Cosmos3OmniTransformer.from_pretrained(
    "nvidia/Cosmos3-Nano", subfolder="transformer", torch_dtype=torch.bfloat16
)

Cosmos3OmniTransformer[[diffusers.Cosmos3OmniTransformer]]

diffusers.Cosmos3OmniTransformer[[diffusers.Cosmos3OmniTransformer]]

Source

forwarddiffusers.Cosmos3OmniTransformer.forwardhttps://github.com/huggingface/diffusers/blob/vr_13921/src/diffusers/models/transformers/transformer_cosmos3.py#L549[{"name": "input_ids", "val": ": Tensor"}, {"name": "text_indexes", "val": ": Tensor"}, {"name": "position_ids", "val": ": Tensor"}, {"name": "und_len", "val": ": int"}, {"name": "sequence_length", "val": ": int"}, {"name": "vision_tokens", "val": ": list"}, {"name": "vision_token_shapes", "val": ": list"}, {"name": "vision_sequence_indexes", "val": ": Tensor"}, {"name": "vision_mse_loss_indexes", "val": ": Tensor"}, {"name": "vision_timesteps", "val": ": Tensor"}, {"name": "vision_noisy_frame_indexes", "val": ": list"}, {"name": "sound_tokens", "val": ": list[torch.Tensor] | None = None"}, {"name": "sound_token_shapes", "val": ": list[tuple[int, int, int]] | None = None"}, {"name": "sound_sequence_indexes", "val": ": torch.Tensor | None = None"}, {"name": "sound_mse_loss_indexes", "val": ": torch.Tensor | None = None"}, {"name": "sound_timesteps", "val": ": torch.Tensor | None = None"}, {"name": "sound_noisy_frame_indexes", "val": ": list[torch.Tensor] | None = None"}, {"name": "action_tokens", "val": ": list[torch.Tensor] | None = None"}, {"name": "action_token_shapes", "val": ": list[tuple[int, int, int]] | None = None"}, {"name": "action_sequence_indexes", "val": ": torch.Tensor | None = None"}, {"name": "action_mse_loss_indexes", "val": ": torch.Tensor | None = None"}, {"name": "action_timesteps", "val": ": torch.Tensor | None = None"}, {"name": "action_noisy_frame_indexes", "val": ": list[torch.Tensor] | None = None"}, {"name": "action_domain_ids", "val": ": list[torch.Tensor] | None = None"}]- input_ids -- Text token IDs placed at text_indexes in the joint sequence.

• text_indexes -- Indices of text tokens in the joint sequence.
• position_ids -- [3, sequence_length] mRoPE position IDs for the full joint sequence.
• und_len -- Length of the causal text (understanding) prefix; generation tokens follow.
• sequence_length -- Total length of the joint packed sequence.
• vision_tokens -- Per-item vision latent tensors before patchify.
• vision_token_shapes -- Patch grid shapes (T, H, W) per vision item.
• vision_sequence_indexes -- Indices of vision tokens in the joint sequence.
• vision_mse_loss_indexes -- Indices used to read vision predictions after the backbone.
• vision_timesteps -- Per-patch diffusion timesteps for vision tokens.
• vision_noisy_frame_indexes -- Noisy frame indices per vision item.
• sound_tokens -- Optional sound latent tensors before packing.
• sound_token_shapes -- Optional patch grid shapes for sound items.
• sound_sequence_indexes -- Optional indices of sound tokens in the joint sequence.
• sound_mse_loss_indexes -- Optional indices used to read sound predictions.
• sound_timesteps -- Optional per-token diffusion timesteps for sound.
• sound_noisy_frame_indexes -- Optional noisy frame indices per sound item.
• action_tokens -- Optional action latent tensors before packing.
• action_token_shapes -- Optional patch grid shapes (T, H, W) per action item.
• action_sequence_indexes -- Optional indices of action tokens in the joint sequence.
• action_mse_loss_indexes -- Optional indices used to read action predictions after the backbone.
• action_timesteps -- Optional per-token diffusion timesteps for action tokens.
• action_noisy_frame_indexes -- Optional noisy frame indices per action item.
• action_domain_ids -- Optional per-item domain IDs selecting the action head weights.0(preds_vision, preds_sound, preds_action) — lists of per-modality predictions. Optional modalities return None when their inputs are omitted. Run a full denoising-step forward pass.

Parameters:

input_ids : Text token IDs placed at text_indexes in the joint sequence.

text_indexes : Indices of text tokens in the joint sequence.

position_ids : [3, sequence_length] mRoPE position IDs for the full joint sequence.

und_len : Length of the causal text (understanding) prefix; generation tokens follow.

sequence_length : Total length of the joint packed sequence.

vision_tokens : Per-item vision latent tensors before patchify.

vision_token_shapes : Patch grid shapes (T, H, W) per vision item.

vision_sequence_indexes : Indices of vision tokens in the joint sequence.

vision_mse_loss_indexes : Indices used to read vision predictions after the backbone.

vision_timesteps : Per-patch diffusion timesteps for vision tokens.

vision_noisy_frame_indexes : Noisy frame indices per vision item.

sound_tokens : Optional sound latent tensors before packing.

sound_token_shapes : Optional patch grid shapes for sound items.

sound_sequence_indexes : Optional indices of sound tokens in the joint sequence.

sound_mse_loss_indexes : Optional indices used to read sound predictions.

sound_timesteps : Optional per-token diffusion timesteps for sound.

sound_noisy_frame_indexes : Optional noisy frame indices per sound item.

action_tokens : Optional action latent tensors before packing.

action_token_shapes : Optional patch grid shapes (T, H, W) per action item.

action_sequence_indexes : Optional indices of action tokens in the joint sequence.

action_mse_loss_indexes : Optional indices used to read action predictions after the backbone.

action_timesteps : Optional per-token diffusion timesteps for action tokens.

action_noisy_frame_indexes : Optional noisy frame indices per action item.

action_domain_ids : Optional per-item domain IDs selecting the action head weights.

Returns:

(preds_vision, preds_sound, preds_action) — lists of per-modality predictions. Optional modalities return None when their inputs are omitted.