Gets the number of unfinished requests.

获取物体应当被放置到的位置 考虑因素： 1. 三维坐标与给定坐标一致 2. 物体的朝向合理 - 物体 z 轴与给定坐标 z 轴一致 - 满足在 xy 平面上的一定约束 -

Return world size for the tensor model parallel group.

Return world size for the tensor model parallel group.

Get ulysses sequence parallel world size.

Returns True if there are unfinished requests.

NOTE: This method is a hack from the open-sourced version without asertion of world_size = tp * pp Initialize model parallel groups.

Set input tensor to be used instead of forward()'s input. When doing pipeline parallelism the input from the previous stage comes fro

control_seq: left_arm, right_arm, left_gripper, right_gripper

TODO(zhangchi.usc1992): We may change the implementation later

将世界坐标系下的矩阵转换到 base 坐标系下

Validate and resolve the unnormalization key for action statistics

Create a TIMM-based featurizer model with appropriate configurations. Args: model_id: The TIMM model ID to load

Returns: entropy: # (bs, response_len) log_probs: # (bs, response_len)

Generate one step of actions

Get Stats to be Logged to Prometheus.

Normalize answer expressions.

Prepares input for action prediction by adding necessary tokens

Creates labels tensor for action prediction if not provided

Prepare the output batch from VLA history

Helper to get action masks from labels

Process vision features with optional FiLM conditioning

transform target pose from world frame to base frame base_pose: np.array([x, y, z, qw, qx, qy, qz]) t

Unnormalize actions using dataset statistics

Add a request to the engine's request pool. The request is added to the request pool and will be processed by the scheduler as `engin

allgather params of all pp ranks. Return a list of handles

Check and sync model logic files between current code and checkpoint. Handles the relationship between current and checkpoint versions of bo

Converts a string to a sequence of ids (integer), using the tokenizer and vocabulary. Args: text (`str`, `List[str]` or

Filter instructions to only include those that have all placeholders matching the available episode parameters. No more, no less. Also ac

Args: input_ids: input ids. shape (batch_size, seq_length) attention_mask: attention_mask. shape (batch_size, seq_le

We assume: - The model takes input: (input_ids, attention_mask, position_ids). No rmpad for the input - The communication sha

Generate descriptions for episodes by replacing placeholders in instructions with parameter values. For each episode, filter instructions tha

Get all the logged statistics for the given dataset.

Get all the parameters of the models in all pp ranks Returns: params: List[List[Dict[str, Tensor]]]: a list of parameters in all

Extracts image from observations and preprocesses it.