Adds a new event to the database.

Adds a new event to the database.

Applies Rotary Position Embedding to the query and key tensors. Args: q (`torch.Tensor`): The query tensor. k (`torch.Tensor`): T

Check if a directory exists. Args: path: The path to check. env: The environment. Returns: Whether the directory exists.

Check if a directory exists. Args: path: The path to check. env: The environment. Returns: Whether the directory exists.

* Returns a function to use in pseudos for positionals * @param {Function} fn

* Returns a function to use in pseudos for positionals * @param {Function} fn

A func to sync attention result with alltoall in sequence parallel gather head dimension and scatter seq dim: e.g. seq_dim: 1, head_dim:

用于构建静态评测

Generate random parameters for a new calendar event task.

Generate random parameters for a new calendar event task.

Return nccl backend type based on the device type. Returns: nccl backend type string.

Gets the task registry for the given family. Args: family: The family. Returns: Task registry. Raises: ValueError: If

Gets the task registry for the given family. Args: family: The family. Returns: Task registry. Raises: ValueError: If

end of sentence token can be int or list: 1 or [1, 2] e.g. response_id = torch.tensor([[20, 10, 34, 1, 0, 0, 0],

Build the inverse of an index mapping. Args: idx_map (Sequence[int]): Sequence where idx_map[i] = j. Returns: List[int]

Load and initialize a reward manager based on the configuration. Args: config: PPO trainer configuration object containing reward_mo

Release the tool instance. Args: instance_id: The instance id of the tool.

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

Saves a task's episodes to disk.

Saves a task's episodes to disk.

Initialize the task.

Initialize the task.

broadcast tensor from rank0 across mp_group

broadcast tensor across mp_group

broadcast tensor from rank0 across mp_group

fetch tensor

fetch tensor

Create a base TransformerConfig with common parameters across different model architectures. TODO: (ycl) use dataclass or converter config?