Args: x (torch.Tensor): (N, D) tensor of spatial positions

x: B, N, C, [resolution, resolution, resolution] cond: B, N, C, H, W

Apply the block to a Tensor, conditioned on a timestep embedding. Args: x: an [N x C x ...] Tensor of features.

Load a model from a pretrained checkpoint. Args: path: The path to the checkpoint. Can be either local path or a Hugging Face model

Load the TRELLIS base models plus IScene release weights.

Get the conditioning information for the model. Args: image (Union[torch.Tensor, list[Image.Image]]): The image prompts.

Copied from Plenoxels Continuous learning rate decay function. Adapted from JaxNeRF The returned rate is lr_init when step=0 and lr_fina

Load one scene RGB image and split a multi-label segmentation into per-instance masks. The segmentation can be single-channel label IDs, pal

A context manager that ensures local master process executes first.

A context manager that ensures master process executes first.

Multi-instance self-attention. q stays (B_total, L, ...). k, v are concatenated across instances (N) -> (B, N*L, ...), then e

Read the binary file distributedly. File is only read once by the rank 0 process and broadcasted to other processes. Returns: da

Render the gausssian. Args: gaussian : gaussianmodule extrinsics (torch.Tensor): (4, 4) camera extrinsics

Generate samples from the model using Euler method. Args: model: The model to sample from. noise: The initia

Generate samples from the model using Euler method. Args: model: The model to sample from. noise: The initia

Scale the parameters of a module and return it.

Broadcast a 1D tensor to a sparse tensor along the batch dimension then perform an operation. Args: input (torch.Tensor): 1D tensor

Concatenate a list of sparse tensors. Args: inputs (List[SparseTensor]): List of sparse tensors to concatenate.

Convert a generated asset to a glb file. Args: app_rep (Gaussian): Appearance representation. mesh (MeshExtractResult): Extr

Unwrap the model from distributed training.