Append values from new_data to lists in data. For each key in new_data, this function appends the corresponding value to a list stored under

In megatron, the model and optimizer storage are: - bf16 parameter data chunked in model parallel group - fp32 grad chunked in model para

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

Get ulysses sequence parallel process group.

Check if the OCR tool is enabled and ready to use. Returns: True if the tool is enabled, False otherwise

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

Gather a tensor across a process group and optionally unpad its padded elements. Args: x (Tensor): Input tensor to gather. g

Compute the mean of `values` over elements selected by `mask`. Args: values (Tensor): Input tensor. mask (Tensor): Boolean o

Spawn to a dictionary of worker groups, each with a subset of method with prefix. Args: prefix_set: Set of prefixes to create wor

Pad and slice input_ids to be divisible by sp_size Pad position_ids to be divisible by sp_size. Note both input_ids_rmpad and position_i

Close the environment and release resources.

Create a DataProto from a dict of tensors and non_tensors

Softmax with mask (optional)

Calculate global rank with support for CP/EP parallelism

Clear the cache and reset statistics.

Encode an observation

Tell the agent that it's evaluation phase.

Build the model.

Load a external data type based on the file path and type name

Partially borrowed from Keras # Arguments sequences: list of lists where each element is a sequence maxlen: int, maximum leng

Wrapper for `receive` handler for performing search action on current page

Wake up engine to load model weights and build kv cache.

broadcast tensor from rank0 across mp_group

Apply monkey patch to the models for ulysses sequence parallel and fused kernel. In the end of this function forward function of the model i

Generate an action by maximum q values.

Function that gets the torch.device based on the current machine. This currently only supports CPU, CUDA, NPU. Returns: device

Create a huggingface processor to process multimodal data. Args: name_or_path (str): The name of the processor. Returns: tra

Log a value of some diagnostic Call this once for each diagnostic quantity, each iteration If called many times, last value will be used.

Get the rank of this worker in the distributed setup.

Reduces a dictionary of metric lists by computing the mean, max, or min of each list. The reduce operation is determined by the key name:

Select specific indices from the DataProto. Args: idxs (torch.Tensor or numpy.ndarray or list): Indices to select

given TP,DP,PP rank to get the global rank.

given TP,DP,PP rank to get the global rank.

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

Create a random mask given input_ids. Support left padding and right padding. Process: - Sample valid token length - Sample left_padding l

Create a dataset. Arguments: data_config: The data config. tokenizer (Tokenizer): The tokenizer. processor (Processor): T

Fetches memory records at a specific time step across all environments.

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:

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

Compute per-token log-probabilities for the given labels. Uses a Flash-Attention–based cross-entropy (if available) for efficient backward,

Creates a batch generator suitable for Megatron pipeline parallelism, handling virtual pipeline parallelism (VPP). If VPP is used (vpp_s