Update rendering to refresh the camera's RGBD information (rendering must be updated even when disabled, otherwise data cannot be col

Convenience method to convert pymunk coordinates to pygame surface local coordinates. Note that in case positive_y_is_up is False, this funct

Helper method to apply gradient checkpointing if enabled.

Returns a bytes_list from a string / byte.

Get the transformation matrix of given contact point of the actor.

This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, num_key_value_heads, seqlen, he

Set gripper posture - `left_pos`: Left gripper pose - `right_pos`: Right gripper pose - `set_tag`: "left" to set the

This method forwards all its arguments to Qwen2TokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please refer to the docstring o

Update all metrics in `self.state` by iterating through special positional arguments & kwargs.

Get the point of the entity actor.

Create a model with the given parameters.

Load a 3D model from a GLB file. Args: file_path: Path to the .glb file Returns: trimesh.Scene object

Start the model server and listen for incoming client connections

Append a new message.

This method defines the mapping between unstructured data (i.e., nested dict) to the structured Observation format.

Returns the freeze filter based on the model config.

Apply function @func to torch.Tensor objects in a nested dictionary or list or tuple. Args: x (dict or list or tuple): a possibl

Apply function @func to torch.Tensor objects in a nested dictionary or list or tuple. Args: x (dict or list or tuple): a possibl

Resize an image to match the policy's expected input size. Uses the same resizing scheme as in the training data pipeline for distribution m

Calculate the center points of each face of the bounding box. Returns: List of face center coordinates

Check if two actors are in contact. - actor1: The first actor. - actor2: The second actor.

Set the scene - Set up the basic scene: light source, viewer.

Returns continuous actions for discrete action token IDs. NOTE =>> Because of the way the actions are discretized w.r.t. the bins (a

Flatten a nested dictionary. Uses '/' as the separator.

Flatten a nested dictionary. Uses '/' as the separator.

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

Returns the freeze filter based on the model config.

Obtain the grasp pose through the marked grasp point. - actor: The instance of the object to be grasped. - arm_tag: The arm t