Compute and return the statistics of the vectors processed so far. Returns: dict: A dictionary containing the computed s

Load and initialize the VLA model from checkpoint. Args: cfg: Configuration object Returns: torch.nn.Module: The initia

Generate action predictions with the VLA policy. Args: cfg: Configuration object with parameters vla: The VLA model

Loads the model weights. Args: params: Parameters of the model. This is a nested structure of array-like objects that

Copied from big_vision. Tokens can attend to valid inputs tokens which have a cumulative mask_ar smaller or equal to theirs. This way `mask_a

Adapted from big_vision. Tokens can attend to valid inputs tokens which have a cumulative mask_ar smaller or equal to theirs. This way `mask_

Adapted from big_vision. Tokens can attend to valid inputs tokens which have a cumulative mask_ar smaller or equal to theirs. This way `mask_

This function is responsible for loading a specific RLDS dataset from storage and getting it into a standardized format. Yields a dataset of

Environment for simulated robot bi-manual manipulation, with end-effector control. Action space: [left_arm_pose (7), # posit

Environment for simulated robot bi-manual manipulation, with joint position control Action space: [left_arm_qpos (6), # abso

Plan a batch of trajectories for multiple target poses. Input: - curr_joint_pos: List of current joint angle

lang_tokens: (batch_size, lang_len, lang_token_dim) lang_attn_mask: (batch_size, lang_len), a mask for valid language tokens,

Given a batch of last hidden Transformer layer embeddings (which fuse the vision-language observation embeddings, noisy action embedd

Append transforms to the group and return a new group. Args: inputs: Appended to the *end* of the current input transforms.

Render a view from a specified position in the Polaris system, with camera direction calculated as position-to-centroid vector.

Render the scene from a specific camera position pointing in a specific direction. Args: camera_position: Position of the

Reset the agent to its initial state.

Reset the agent to its initial state.

Takes numpy array corresponding to a single image and resizes to 256x256, exactly as done in the ALOHA data preprocessing script, which is us

Rotates half the hidden dims of the input.

To be called *between* epochs, prior to DataLoader instantiation; ensures random order across epochs.

control_seq: left_arm, right_arm, left_gripper, right_gripper

Converts all numpy arrays in nested dictionary or list or tuple to torch tensors (and leaves existing torch Tensors as-is), and returns a

将世界坐标系下的矩阵转换到 base 坐标系下

Initializes FiLM ViT block wrapper. Args: block (timm.models.vision_transformer.Block): Vision transformer block.

Extract the first or third Euler angle from the two members of the matrix which are positive constant times its sine and cosine. Args:

Applies RoPE positions [B, L] to x [B, L, H, D].

Applies RoPE positions [B, L] to x [B, L, H, D].

Applies RoPE positions [B, L] to x [B, L, H, D].

Applies RoPE positions [B, L] to x [B, L, H, D].

Validate and resolve the unnormalization key for action statistics