r"""Build and validate the evaluation arguments.

r"""Parse eval example. input: a dict with keys {"question", "A", "B", "C", "D", "answer"} output: a tuple of (prompt, response).

r"""Build and validate the training arguments.

r"""Pre-process a single image.

r"""Preview the training commands.

r"""Asynchronously get a list of responses of the chat model.

r"""Asynchronously get a list of scores of the reward model.

r"""Asynchronously get the response token-by-token of the chat model.

Batch data into lists of length n. The last batch may be shorter.

r"""Calculate effective tokens per second.

r"""Compute the sum log probabilities of the labels under given logits if loss_type is not IPO, ORPO or SimPO. Otherwise the average log prob

r"""Return the number of trainable parameters and number of all parameters in the model.

r"""Extract tool message.

r"""Find the modules in the expanded blocks to apply lora.

r"""Run forward pass and computes the log probabilities.

r"""Compute the log probabilities of the given labels under the given logits. Returns: logps: A tensor of shape (batch_size,) containing

r"""Convert dataset or dataset dict to dataset module.

Count number of words in a tokenized line (words longer than 2 characters). Args: line: List of word tokens Returns

r"""Return the ollama modelfile. TODO: support function calling.

Generate standardized rewrite method name for storing results. Args: dataset: Name of the dataset (Researchy-GEO, E-commerce, or GEO-

Generate prompt for merging rules. Args: rules_list: List of rules to merge Returns: Formatted prompt stri

r"""Get the sequnce lengths in the current batch. e.g. ```python # input [ [1, 1, 2, 2, 2, 0], [1, 2, 2, 3, 3, 3],

Calculate and save GEU scores for documents. Args: num_examples: Number of examples to process (None for all) engine_llm: LLM

r"""Infer the optimal dtype according to the model_dtype and device compatibility.

r"""Load the training configuration from config path.

r"""Load value head parameters from Hugging Face Hub or local disk. Returns: dict with keys `v_head.summary.weight` and `v_head.summary.bias`.

Log progress information. Args: current: Current progress total: Total items item_name: Name of i

Make a request with retry logic for rate limiting.

r"""Detach `tensors` (even if it's a nested list/tuple/dict of tensors).

Basic string normalization.

r"""Expand 2d attention mask to 4d attention mask. Expand the attention mask with indices from (batch_size, seq_len) to (batch_size, 1, seq_len,

r"""Get arguments from the command line or a config file.

Reconstruct dataset split into multiple JSON files in a directory. Args: dataset_split: Dataset split from Hugging Face datasets

r"""Replace the default/reward modules in the model. The model is already unwrapped.

Execute multiple scripts using MorphCloud via the router. Args: scripts: List of code scripts to execute. la

r"""Save the training configuration to config path.

Scores all test cases in a subtask. Args: client: PistonClient instance for executing code subtask: Dictionary containing su

r"""Set up a filesystem object based on the path protocol.

r"""EMA implementation according to TensorBoard.

Converts the SubtaskResult to a dictionary representation. Returns: dict: Dictionary containing all subtask result data

Generate word-level n-grams from text.

r"""Init a runner.

Compile the code on the instance. Args: instance: The MorphCloud instance Returns: InstanceExecResp

Internal implementation of execute with no retry logic. Args: data: Dictionary containing execution data Return

Execute code using a prepared instance. Args: instance: Ready MorphCloud instance data: Execution data te

Determines the status code based on the score and feedback message. Args: score: The numeric score (0.0 to 1.0) feedback: Th

Filter rules to remove query-specific dependencies and keep only general principles. Args: rules_list: List of rules to filter

r"""Clean the cached memory and resets the runner.

Fix common JSON issues in LLM responses, especially unescaped quotes. Args: text: JSON string that may have formatting issues

r"""Post-process the response text. Based on: https://huggingface.co/spaces/Lyte/DeepSeek-R1-Distill-Qwen-1.5B-Demo-GGUF/blob/main/app.py

Get or create Anthropic client. Returns: Anthropic client instance Raises: ValueError: If ANTHROPIC_API_KEY not

Get the compile script content.

r"""Return the corresponding dataset processor.

r"""Return the default logging level.

r"""Get gemma3 token type ids for computing loss. Returns: batch_token_type_ids: shape (batch_size, seq_length)

r"""Return the ollama template.