Toggle skill usage (for w/ vs w/o skill validation).

Decorator to add a timeout to a function. Args: seconds: The timeout duration in seconds. use_signals: (Deprecated) This is

Split along the second dim into `n_split`, unfold it to the first dim (batch dim) Useful in passing grouped tensors that doesn't want to be sh

Union two tensordicts.

Fast text wrapping based on character count. Returns a list of tuples (line_text, is_paragraph_end) to track paragraph boundaries. Optimi

Precise text wrapping using actual font measurements for optimal packing. Returns a list of tuples (line_text, is_paragraph_end) to track par

Build a Segment Tree data structure. https://en.wikipedia.org/wiki/Segment_tree Can be used as regular array, but with two

broadcast tensor in tp shards across mp_group

broadcast tensor in tp shards across mp_group

broadcast tensor in tp shards across mp_group

broadcast tensor in tp shards across mp_group

Creates the train and validation dataloaders.

fetch tensor in tp shards across mp_group

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

Helper function containing the loading hf model logic

If curriculum_learning is enabled, build CurriculumSkillManager and inject skills.

Normalize answer expressions.

Render compact mode image using cached complete lines. Args: env_idx: Environment index context: Cur

Render a line of text with multiple highlighted contexts in different colors. Uses cumulative prefix width calculation to ensure proper

Run validation. When curriculum learning is enabled, runs twice (with/without skills) and updates skill set.

Externally exposed constructor function to create parallel Gym environments. - env_name: [gym_cards/Blackjack-v0, gym_cards/NumberLine-v0, gy

This function builds the text observation for the agent and optionally renders trajectory history as images. Returns:

This function builds the text observation for the agent and optionally renders trajectory history as images. Returns:

This function builds the text observation for the agent.

This function builds the text observation for the agent.

This function builds the text observation for the agent.

Mirror *build_sokoban_envs* so higher‑level code can swap seamlessly.

Bulk offload group.

Calculate a value based on majority voting. This function identifies the most common value for a specified vote key in the data, then re

Calls the remote sandbox API to execute code with retry logic for Gateway Timeout, using increasing delay between retries. Logs internal call

Calls the remote search API to perform retrieval with retry logic for various errors, using increasing delay between retries. Logs internal