Puts a list of data into the Ray object store in parallel using a thread pool. Args: data_list (List[Any]): A list of Python objects

Prepare a batch for dynamic batching. Args: data (DataProto): The input data. max_token_len (int): The maximum token length

Register a successfully saved checkpoint and enforce retention limit. Adds the new checkpoint path to tracking and removes excess ol

Move model parameters, optimizer states, or both to the specified device. Note that this function executes irrespective of offload co

Calculate the aligned size of a torch tensor If the tensor's size (in bytes) cannot be evenly divided by _ALIGN_SIZE, it will be rounded

broadcast tensor from rank0 across mp_group

broadcast tensor across mp_group

broadcast tensor from rank0 across mp_group

fetch tensor

fetch tensor

转发请求到指定的worker

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

返回某个会话在 Redis 中的 Hash key

Automatically configure device name for different accelerators. For example, on Ascend NPU, this function defaults the trainer device to "npu"

calculate rollout vs actor logprobs diff, for debugging purpose Args: data: DataProto the data batch to calculate

Compute logits given a batch of data. Args: data (DataProto): a batch of data represented by DataProto. It must contain key ```in

计算a◎b对应的复数平方

Convert Hydra configs and other special types to regular Python types.

统计文件行数

Generates a memory snapshot and saves it as a pickle file in a specified directory. The files are organized by timestamp in subdirect

Return the most recent checkpoint directory based on a tracker file. Args: path (str): Base directory containing the checkpoint trac

Generate text using the SGLang server. Args: prompt (Optional[str], optional): Text prompt for generation. Defaults to None.

Generate tokens from prompt ids. Args: request_id (str): request id for sticky session. prompt_ids (List[int]): List

Get FSDP wrap policy for the module. Args: module: The module to get wrap policy for config: Configuration for wrap policy

Calculates partitions of indices from seqlen_list such that the sum of sequence lengths in each partition is balanced. Uses the Karmarkar-Kar

custom_graph: {"room1": {"exits": {"room2": "direction2", ...}}, ...}

Get the environment variable name for visible device selection. Returns the appropriate environment variable name based on the available acce

Init standalone rollout server, create new resource pool for this rollout.

Check if a path is a non-local (HDFS) path. Args: path (str): The path to check. Returns: bool: True if the path is an HDFS

递归地将列表转换为元组

Convert NestedTensor from no-padding to right padding format. Args: nested_tensor: NestedTensor with no-padding format data:

Generate data using the provided data generation strategy. Note: This method is intended to change the dataset after each training ba

Pad a DataProto to size divisible by size_divisor Args: size_divisor (int): size divisor Returns: data: (DataProto): the pad

pad a 2D tensors (e.g. responses, logprobs) in the last dim to max_seq_length. input shape: [bs, seq_length] output shape: [bs, max_seq_l

Applies a monkey patch to the forward method of a given model class to enable Ulysses sequence parallelism input slicing.

convert point in sketch plane local coordinates to global coordinates

Postprocess packed sequences

Release GPU memory occupation temporarily. Args: tags (Optional[List[str]], optional): List of tags to specify which memory to re

在可滚动容器中渲染 Markdown，使用户可自定义滚动

Note that this operation is in-place

Resume rollout weights or kv cache in GPU memory. Args: tags: weights or kv_cache.

Shutdown the HTTP server and clean up resources. This method performs the following cleanup operations: 1. Unregisters the worker fro

完整实现的解题算法

Move model parameters, optimizer states, or both to the specified device. Note that this function executes irrespective of offload co

Unpad the data proto with pad_size. i.e. `data[:-pad_size]`

Update weights from trainer to rollout replicas.

Update model weights from tensor data. The HTTP server will only post meta data, and the real weights will be copied directly from GP

Validate actor configuration with runtime parameters.

broadcast tensor across mp_group

broadcast tensor across mp_group

内部方法：检查并删除过期的环境

内部方法：必须在已持有锁的上下文中调用